Timeline
1990 outputs, 536 news ·
| Date ▾ | Name | Lab | Type | Stars | Downloads | Citations | Params | Active | Tokens | Context | Intel | Questions | Tasks |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026-11-16 | ★ Qwen4-Max | Alibaba | model | 2.4T | 71B | 1M | 43 | ||||||
| 2026-11-16 | Qwen4-Max Launches: 2.4T-A71B MoE on the New Qwen4 Architecture (QSA + Gated DeltaNet, Gated Residual, 328B N-gram Embeddings), 1M Context, $1.50/$4.50; AA Intelligence Index 43; Smaller Qwen4 Models Coming to Open Source Qwen | Alibaba | news | ||||||||||
| 2026-11-09 | ★ Claude Fable 5.2 | Anthropic | model | 1M | 65 | ||||||||
| 2026-11-09 | ★ Claude Mythos 5.2 | Anthropic | model | 1M | |||||||||
| 2026-11-09 | ★ Muse Ember | Meta | model | 1M | 36 | ||||||||
| 2026-11-09 | Anthropic Reportedly Plans to Set an IPO Valuation Range This Week and Begin Trading Next Week, as Some Investors Push Below the $1.5T Target (WSJ) TechCrunch | Anthropic | news | ||||||||||
| 2026-11-09 | Claude Fable 5.2 and Mythos 5.2 Released: $8/$40 per MTok, 67.6% on Terminal-Bench 4.0, AA Intelligence Index v5.0.1 65 (#1) Anthropic | Anthropic | news | ||||||||||
| 2026-11-09 | ByteDance Releases Seedream 5.5 Image Model in Doubao and Jimeng: Faster Than Seedream 5.0 Pro, Pixel-Level Preservation Over 100+ Edit Rounds, Native RGBA Transparency, at Under Half the Price; Pro Tier to Follow Leiphone | ByteDance | news | ||||||||||
| 2026-11-09 | Google Adds TPU 8tx (Confidential Inference) and TPU 8i Max (Two 8i Chips per Package, Double the Compute, SRAM and HBM Bandwidth), Its Third and Fourth TPU Systems of 2026 Google Cloud | news | |||||||||||
| 2026-11-09 | LG AI Research Opens Free API Access to K-EXAONE 2.5, a Post-Trained Update of K-EXAONE 2.0 Developed With NVIDIA Nemotron Datasets, Claiming a 70%+ Gain Over the August Model; K-EXAONE 3.0 Targeted for January 2027 Seoul Economic Daily | LG | news | ||||||||||
| 2026-11-09 | Meta Introduces Muse Ember, Its Smallest and Cheapest Muse Model and the First Trained Using Muse Wonder Full: 1M Context, $0.10/$0.50 per M Tokens, AA Intelligence Index 36 at $0.14 per Task Meta | Meta | news | ||||||||||
| 2026-11-09 | NVIDIA Plans to Invest in Motif Technologies at the MSIT–NVIDIA AI Startup Partnership Forum, Joining a Bridge Round for Motif 4 (Open Weights Targeted for Q1 2027); Amount and Valuation Undisclosed NVIDIA Blog | Motif Technologies | news | ||||||||||
| 2026-11-09 | OpenAI Describes Control Room, a Central System to Track Model Workloads Across Its Research Infrastructure and Rapidly Stop Them Over Safety or Security Concerns OpenAI | OpenAI | news | ||||||||||
| 2026-11-09 | Sarvam Names Former Google Cloud India MD Sashikumar Sreedharan President and Chief Business Officer, With a Board Seat, to Lead Its Enterprise and Government Business Sarvam AI | Sarvam | news | ||||||||||
| 2026-11-09 | NVIDIA Plans to Invest in Upstage's Ongoing Second Series C Close, Announced at the MSIT–NVIDIA AI Startup Partnership Forum; Joins Amazon and AMD as Strategic Investors, Amount and Stake Undisclosed ZDNet Korea | Upstage | news | ||||||||||
| 2026-11-08 | ★ InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure | TokenRhythm | paper | ||||||||||
| 2026-11-08 | NVIDIA Board Opens an Independent Investigation After Reports That Jensen Huang Was the Target of a Foreign Intelligence Effort; Huang Tells Staff He Was Unfaithful to His Wife and Will Cooperate NVIDIA | NVIDIA | news | ||||||||||
| 2026-11-08 | InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure (1.83x training speedup, open source) TokenRhythm | TokenRhythm | news | ||||||||||
| 2026-11-07 | Agnes 2.5 Pro Theta | Sapiens AI | model | 1M | 24 | ||||||||
| 2026-11-06 | China Mobile Cloud Signs Cooperation Agreement With Zhipu: Joint Token Development Center, GLM Models on the MoMA MaaS Platform, Co-Built AI Coding Product IT之家 | China Mobile (Jiutian) | news | ||||||||||
| 2026-11-06 | OpenAI Closes an Initial $11B of a New Round at a $1.25T Valuation, Short of the Original Up-to-$30B Target; SoftBank and MGX Participate (Reuters) Reuters via The Star | OpenAI | news | ||||||||||
| 2026-11-06 | Reflection Publishes the Beam Model Technical Report: 501.3B-A23.4B MoE, 23.8T Tokens, Depth-Scaled Residuals, and a 100M-Rollout RL Run It Says Had Not Plateaued Reflection AI | Reflection AI | news | ||||||||||
| 2026-11-06 | Zhipu Signs Cooperation Agreement With China Mobile Cloud: Joint Token Development Center, GLM Models on the MoMA MaaS Platform, Co-Built AI Coding Product IT Home | Z.ai | news | ||||||||||
| 2026-11-05 | ERNIE 5.5 Flash Base | Baidu | model | 280B | 14B | 1.05M | |||||||
| 2026-11-05 | PAN 2 | MBZUAI | model | 29B | |||||||||
| 2026-11-05 | ★ Inkling 2 Small | Thinking Machines | model | 594B | 17B | 1M | 31 | ||||||
| 2026-11-05 | ZYFA: Speed and Accuracy of FP8 FlashAttention on AMD CDNA4 | Zyphra | paper | ||||||||||
| 2026-11-05 | Baidu Open-Sources ERNIE 5.5 Flash Base: 280B-A14B MoE Base Model Pre-Trained on 33.5T Tokens With Muon in FP8, Post-Trained Flash to Follow Baidu ERNIE | Baidu | news | ||||||||||
| 2026-11-05 | openPangu-2.1-Pro Appears First in Huawei's Xiaoyi Work; Huawei Cloud MaaS Update Set for November 15 IT之家 | Huawei | news | ||||||||||
| 2026-11-05 | Mistral Acquires SkipLabs, a Paris/London AI Code-Generation Startup Founded by Ex-Meta Engineer Julien Verlaguet, Whose Team Joins Guillaume Lample's Science Team; Its Fifth Acquisition of 2026 After London Design Studio New Material in October (Terms Undisclosed) Maddyness | Mistral | news | ||||||||||
| 2026-11-05 | CNBC: Moonshot Has Discussed Leasing Nvidia Blackwell Clusters of 8,000+ GPUs From European Neoclouds, Mainly for Inference, at a Reported 30-50% Upfront Premium; No Deal Confirmed CNBC | Moonshot AI | news | ||||||||||
| 2026-11-05 | Nanbeige has released Nanbeige4.3-33B, scoring 20 on the Intelligence Index Artificial Analysis | Nanbeige | news | ||||||||||
| 2026-11-05 | OpenAI Releases GPT-6 Cyber, an RL-Post-Trained Cybersecurity Variant of GPT-6 Astra for Approved Daybreak Red Organizations ($20/$100 per MTok) OpenAI | OpenAI | news | ||||||||||
| 2026-11-05 | Thinking Machines Releases Inkling 2 Small: 594B-A17B MoE, AA Intelligence Index 31, Tinker-First With Apache 2.0 Weights Promised Next Week Thinking Machines | Thinking Machines | news | ||||||||||
| 2026-11-04 | GPT-Image-2.6 (Flare and Sunburst) | OpenAI | model | ||||||||||
| 2026-11-04 | ★ Hy4 | Tencent | model | 770B | 35 | ||||||||
| 2026-11-04 | HyScale: Scaling Long-Context Training for Learned Sparse Attention | Tencent | paper | ||||||||||
| 2026-11-04 | Ai2 and Google Launch NeuralGCM+: Multimillion-Dollar, Three-Year Google.org Grant to Build an Open Coupled Atmosphere-Ocean-Sea-Ice Hybrid Climate Model on Google Research's NeuralGCM Ai2 | Ai2 | news | ||||||||||
| 2026-11-04 | Meituan Open-Sources LongCat-2.5 Weights Under MIT: 1.6T-Total / ~48B-Active Native Multimodal MoE, Plus a W4A4 Checkpoint for INT4 on Domestic Accelerators HuggingFace | Meituan | news | ||||||||||
| 2026-11-04 | OpenAI to Acquire Pierre Computer Company, Maker of the Code Storage Git Layer and the Trees and Diffs Libraries Used in Codex; Terms Undisclosed OpenAI | OpenAI | news | ||||||||||
| 2026-11-04 | OpenAI Releases Images 2.6: GPT-Image-2.6 Sunburst and Flare Add Complexity Control, Animated GIFs and SVG Output OpenAI | OpenAI | news | ||||||||||
| 2026-11-04 | Tencent Releases Hy4, Adding Vision and Computer Use, and Says It Ranks Among China's Best Models; Weights to Be Open-Sourced Later Bloomberg | Tencent | news | ||||||||||
| 2026-11-03 | Huawei Cloud Stops Selling openPangu-2.0-Flash and Pro on November 6, Full Retirement November 30; Migration Advice Points to DeepSeek-V4.1-Flash and GLM-5.3 IT之家 | Huawei | news | ||||||||||
| 2026-11-03 | Korea Sovereign AI Project Weighs a Double-Blind User Evaluation for Its Third Stage, Hiding Developer and Model Names; Upstage, SKT and LG AI Research Refine Models Through January 2027, With the Evaluation Due to Conclude Around Mid-February ZDNet Korea | LG | news | ||||||||||
| 2026-11-03 | Meta Seeks to Raise $20B-$25B in a Bond Sale (Two-Thirds Dollars, One-Third Euros) as AI Spending Jumps; S&P Revises Outlook CNBC | Meta | news | ||||||||||
| 2026-11-03 | OpenAI Appoints Jill Hazelbaker as Chief Corporate Affairs Officer, Leading Global Communications OpenAI | OpenAI | news | ||||||||||
| 2026-11-03 | SenseTime Releases SenseNova 6.8 Flash (Official), Adding Cross-Session Long-Term Memory and Dynamic PPT, Poster, Web, SVG and Blender 3D Design IT Home | SenseTime | news | ||||||||||
| 2026-11-03 | Korea Sovereign AI Project Weighs a Double-Blind User Evaluation for Its Third Stage, Hiding Developer and Model Names; Upstage, SKT and LG AI Research Refine Models Through January 2027, With the Evaluation Due to Conclude Around Mid-February ZDNet Korea | SK Telecom | news | ||||||||||
| 2026-11-03 | Korea Sovereign AI Project Weighs a Double-Blind User Evaluation for Its Third Stage, Hiding Developer and Model Names; Upstage, SKT and LG AI Research Refine Models Through January 2027, With the Evaluation Due to Conclude Around Mid-February ZDNet Korea | Upstage | news | ||||||||||
| 2026-11-02 | LongCat-2.5 | Meituan | model | 1.6T | 48B | 1M | |||||||
| 2026-11-02 | LaBraM at Scale: Scaling Data and Models for EEG Pretraining | Shanghai Jiao Tong University, Microsoft | paper | ||||||||||
| 2026-11-02 | ZHA: Zyphra Hierarchical Attention | Zyphra | paper | ||||||||||
| 2026-11-02 | Meituan Makes LongCat-2.5 Generally Available on the LongCat API: Native Multimodal Agentic Model, 1M Context Meituan LongCat | Meituan | news | ||||||||||
| 2026-11-02 | Mistral Large 4 Technical Report: Model Surgery Removed Its KV-Cache Compression After It Hurt Agentic Performance; DeepSeek Sparse Attention, NVFP4 on Four GB300s Hugging Face | Mistral | news | ||||||||||
| 2026-11-02 | Moonshot Releases Kimi K3.1 Pro Open Weights on HuggingFace, Two Days Ahead of Its Nov 4 Promise, Under the Kimi K3.1 Pro License HuggingFace | Moonshot AI | news | ||||||||||
| 2026-11-02 | OpenAI Introduces LifeVal, a Benchmark for How Well AI Helps People Complete Everyday Tasks Such as Finding Childcare, Managing Finances and Government Paperwork OpenAI | OpenAI | news | ||||||||||
| 2026-11-01 | ★ 801-30B-A3B | SII | model | 30B | 3B | 11.8T | 262.14K | ||||||
| 2026-10-30 | SenseNova-Loop-9B-Preview | SenseTime | model | 9B | 262.14K | ||||||||
| 2026-10-30 | Cohere Advances a C$4B R&D and AI-Infrastructure Commitment in Canada Through 2031, With 1,500+ New Roles and a C$309M GPU Cloud Partnership With Bell AI Fabric Cohere | Cohere | news | ||||||||||
| 2026-10-30 | OpenAI Releases GPT-Rosalind-6 for Approved Life-Sciences Researchers: Same Weights as GPT-6 Astra With Fewer Dual-Use Biology Refusals OpenAI | OpenAI | news | ||||||||||
| 2026-10-30 | OpenAI Will Honor a Shared robots.txt Signal for Website Operators' AI Training Preferences OpenAI | OpenAI | news | ||||||||||
| 2026-10-29 | Multi-View Latent MoE: Letting Every Router Shape Expert Computation | Baidu | paper | ||||||||||
| 2026-10-29 | Axion 1.0: Open Agent Foundation Models, Simplified | ByteDance | paper | ||||||||||
| 2026-10-29 | Training Foundation Models with a Single Algorithm | CMU | paper | ||||||||||
| 2026-10-29 | Sarashina3.2 mini thinking | SB Intuitions | model | ||||||||||
| 2026-10-29 | Cohere Announces the Initial Close of a Series E Targeting Up to C$7B, Led by Schwarz Group With Bell Ventures, Bullhound, Inovia, NVIDIA, Prosperity7 and the Government of Canada (a Staged C$1.6B Stake); BetaKit Reports Nearly C$36B Post-Money; AISLE Acquisition and Aleph Alpha Combination Still Pending Cohere | Cohere | news | ||||||||||
| 2026-10-29 | Mistral Releases Mistral Large 4 Weights (1T-A52B) Under a New Modified-MIT Ecosystem License After Three More Weeks of RL Mistral AI | Mistral | news | ||||||||||
| 2026-10-29 | Post-Training in the Open: Marin Snowball 67B-A2B v0.1 Reaches GPT-OSS-Level on Marin's Eval Policy, From a Lineage of 100+ Post-Trained Checkpoints Open Athena | Open Athena | news | ||||||||||
| 2026-10-29 | OpenAI Publishes an Updated Model Spec Built Around Five Core Values for More Capable AI OpenAI | OpenAI | news | ||||||||||
| 2026-10-29 | OpenAI Describes Its Layered Monitoring for Detecting Frontier Model Misalignment and Unauthorized Agent Behavior OpenAI | OpenAI | news | ||||||||||
| 2026-10-29 | Zenrin and Zenrin Datacom Form a Business and Capital Alliance With PFN, Subscribing to a Third-Party Allotment of New PFN Shares to Build Geospatial AI Services (Amount Undisclosed) Preferred Networks | PFN | news | ||||||||||
| 2026-10-29 | SB Intuitions Releases Sarashina3.2 mini thinking, a From-Scratch Reasoning Model With Low/Medium/High Reasoning Effort, Added to the Sarashina Service SB Intuitions | SB Intuitions | news | ||||||||||
| 2026-10-29 | Tencent Says the General Release of Hy4 and a Cheaper Hy4 Flash Are Coming Soon, Both Adding Visual Understanding TechNode | Tencent | news | ||||||||||
| 2026-10-28 | ★ Gemini 4 Carbon | model | 1M | 60 | |||||||||
| 2026-10-28 | Nemotron 3.5 Transcribe | NVIDIA | model | 65.54K | |||||||||
| 2026-10-28 | Anthropic Expands AI for Science to Foundations and Research Institutions Anthropic | Anthropic | news | ||||||||||
| 2026-10-28 | Google Announces Gemini 4 Carbon: #1 on AA Intelligence Index v5.0 (59), SOTA SEC-bench Pro 88.2% and ExploitGym 54.8%, Same $2/$10 Introductory Pricing, Cyber Defenders First via Fairwind Google | news | |||||||||||
| 2026-10-28 | NVIDIA Releases Nemotron 3.5 Transcribe: Open 34B-A4.3B Speech-Augmented LM on Nemotron 3.5 Lightning, 40 Languages, Hour-Long Recordings, OpenMDW-1.1 NVIDIA | NVIDIA | news | ||||||||||
| 2026-10-28 | Cognizant in Talks to Invest in Sarvam's ~$300M Round at a Reported $3.5–4B Valuation, More Than Double June's $1.5B (ET Report; Talks Ongoing) Inc42 | Sarvam | news | ||||||||||
| 2026-10-28 | MatVerse (材华): Suzhou National Laboratory, Shanghai AI Lab and SJTU Launch a Natively Multimodal Materials Foundation Model With the 19-Subfield OmniMatBench Shanghai AI Laboratory | PJLab | news | ||||||||||
| 2026-10-27 | Unlimited-OCR-2 | Baidu | model | ||||||||||
| 2026-10-27 | ★ Genie 4 | model | |||||||||||
| 2026-10-27 | ★ LFM3 | Liquid AI | model | 2.17B (max) | 22T | 262.14K | |||||||
| 2026-10-27 | Can Discrete Tokens Serve as Semantically Lossless Representations of Visual Signals? (FVQ) | Meituan | paper | ||||||||||
| 2026-10-27 | MiniCPM-OCR | OpenBMB | model | 2B (max) | |||||||||
| 2026-10-27 | PLaMo 3.1 Prime | PFN | model | ||||||||||
| 2026-10-27 | inclusionAI Releases Ling-3.1-flash-1019 Weights (MIT): Post-Training Upgrade of the 560B-A25B MoE, AA Intelligence Index 44 on v4.3.2 and 29 on v5.0; Ling-3.1-flash-base Published Two Days Earlier HuggingFace | Ant Group | news | ||||||||||
| 2026-10-27 | Anthropic Reports Early Progress on Accelerating Drug Development With Claude Anthropic | Anthropic | news | ||||||||||
| 2026-10-27 | Google DeepMind Announces Genie 4: Real-Time Interactive World Model With Native Audio and Dialogue, Natural-Language Steering; Research Preview for Trusted Testers Google DeepMind | news | |||||||||||
| 2026-10-27 | Liquid AI Releases LFM3: A 22T-Token Tri-Hybrid 1.7B Base and Text, Vision, Audio, Omni and Decision Models Branched From It Liquid AI | Liquid AI | news | ||||||||||
| 2026-10-27 | NAVER Signs an Up-to-$9B Computing-Infrastructure Agreement With Brookfield (55 MW From H1 2027, 200 MW in 2028), Meeting the Precondition for NVIDIA's Strategic Equity Investment NAVER | Naver | news | ||||||||||
| 2026-10-27 | PFN Releases PLaMo 3.1 Prime, a Hybrid Reasoning/Non-Reasoning Flagship on the NICT Joint-Research Base, With Web Search, Deep Research and ~1.7x Faster Output via Speculative Decoding Preferred Networks | PFN | news | ||||||||||
| 2026-10-27 | Sarvam Signs an MoU With HQ Integrated Defence Staff (Ministry of Defence) on a Joint Framework for AI Across India's Army, Navy and Air Force, Delivered by Its Firewalled Defence Division Chanakya on Models Such as Sarvam 105B Sarvam AI | Sarvam | news | ||||||||||
| 2026-10-26 | GPT-6.1 Terra | OpenAI | model | 1.05M | 51 | ||||||||
| 2026-10-26 | Anthropic Files Public S-1 With the SEC for an IPO of Class A Stock on Nasdaq Under the Ticker ANT; Price Range Not Yet Set Anthropic | Anthropic | news | ||||||||||
| 2026-10-26 | Cohere Reports Rare Unprompted Exfiltration Attempts by Qwen3.8-Flash-Next in Sandboxed Agent Evals, Publishes Traces and Repro Cohere | Cohere | news | ||||||||||
| 2026-10-26 | Meta Prices $37.5B Equity Raise ($25B Common at $625/Share Plus $12.5B Mandatory Convertible Preferred); NVIDIA Commits $3B and Expands Muse Training and Inference Collaboration Meta | Meta | news | ||||||||||
| 2026-10-26 | BOSS 直聘旗下 Nanbeige 开源 Nanbeige4.3-33B (BOSS Zhipin's Nanbeige open-sources Nanbeige4.3-33B) DoNews | Nanbeige | news | ||||||||||
| 2026-10-26 | OpenAI Releases GPT-6.1 Terra: $1/$5 per MTok, 1.05M Context, Critical in Cybersecurity; AA Intelligence Index v5.0 51, Four Points Below GPT-6.1 Sol OpenAI | OpenAI | news | ||||||||||
| 2026-10-25 | AgentBaton: Efficient Model Collaboration in Agentic Tasks | NVIDIA | paper | ||||||||||
| 2026-10-24 | ★ Kimi K3.1 Pro | Moonshot AI | model | 2.8T | 104B | 1M | 46 | ||||||
| 2026-10-24 | ★ Nanbeige4.3-33B | Nanbeige | model | 33B | 524.29K | 20 | |||||||
| 2026-10-24 | MBZUAI and Inception42 Unveil Jais 3, Arabic Models Built on K2 Horizon, and Jais Speech Covering Dialects From 19 Arab Countries The National | MBZUAI | news | ||||||||||
| 2026-10-24 | Moonshot Releases Kimi K3.1 Pro: Post-Trained K3 (2.8T MoE), 1M Context, $2/$10 API; Open Weights Promised by Nov 4 Moonshot AI | Moonshot AI | news | ||||||||||
| 2026-10-23 | Atria Dawn | PJLab | model | 512K | |||||||||
| 2026-10-23 | Volcano Engine Opens Enterprise Trials of Seedance 2.5 4K, a Film-and-TV Tier With 10-Bit Color and 480p Draft Mode TechNode | ByteDance | news | ||||||||||
| 2026-10-23 | OpenAI Details Containment for Frontier Model Training and Evaluation, Including Offline Environments and High-Compute Red-Teaming OpenAI | OpenAI | news | ||||||||||
| 2026-10-22 | ★ Ling-3.1-flash-1019 | Ant Group | model | 560B | 25B | 262.14K | 30 | ||||||
| 2026-10-22 | Anthropic Donates Agent Skills to the Agentic AI Foundation Anthropic | Anthropic | news | ||||||||||
| 2026-10-22 | DeepSeek Reportedly Closes a ~¥100B (~$15B) Round Led by CATL and Tencent Ahead of an IPO; DeepSeek Has Not Confirmed Bloomberg via The Next Web | DeepSeek | news | ||||||||||
| 2026-10-22 | OpenAI in Final-Stage Talks to Raise ~$30B at a $1.25T Pre-Money Valuation, With UAE Investors and Coatue; Not Yet Closed CNBC via The Edge Malaysia | OpenAI | news | ||||||||||
| 2026-10-21 | ★ Raon-1 mini (preview) | KRAFTON | model | 28.8B | 4.3B | 262.14K | |||||||
| 2026-10-21 | Hy-Nomas: Towards Free-form Interactive World Generation | Tencent | model | 8B | |||||||||
| 2026-10-21 | Mistral and NVIDIA Collaborate on AI-Driven Engineering Workflows Linking Simulation, Model Development and Design; Public Physics-Model Benchmarks and Datasets Planned Mistral AI | Mistral | news | ||||||||||
| 2026-10-21 | OpenAI Models and Codex Workflows Come to Google Cloud's Model Garden via Google Cloud Marketplace OpenAI | OpenAI | news | ||||||||||
| 2026-10-20 | OpenAI Acquires Orchestra Bio to Bring AI-Powered Operations to Life-Sciences R&D OpenAI | OpenAI | news | ||||||||||
| 2026-10-20 | PFN and Tenstorrent Selected for a Five-Year NEDO Project to Develop Photonics-Integrated AI Processors, With PFN Building High-Bandwidth 3D-Stacked-Memory AI Chiplets Preferred Networks | PFN | news | ||||||||||
| 2026-10-19 | ★ Mercury 3 | Inception Labs | model | 1M | |||||||||
| 2026-10-19 | ★ MiniMax-M3.1-Flash | MiniMax | model | 23B | 1.05M | 29 | |||||||
| 2026-10-19 | ★ RFD4-Proteína | NVIDIA, University of Washington | model | 436M | |||||||||
| 2026-10-19 | inclusionAI Open-Weights Ling-3.0-flash-Sante (MIT), a Healthcare-Specialized 124B-A5.1B Ling Aimed at Medical Reasoning, Deep Research and Multi-Agent Workflows HuggingFace | Ant Group | news | ||||||||||
| 2026-10-19 | Inception Launches Mercury 3: Diffusion LLM More Than 40x Larger Than Mercury 1, 1M-Token Context, 1,068 Tokens/s; Mercury Voice and Mercury Decide Ship Alongside Inception | Inception Labs | news | ||||||||||
| 2026-10-17 | ECHO: Efficient Causal Self-Recursion Beyond Auxiliary Multi-Token Prediction | Baidu | paper | ||||||||||
| 2026-10-17 | ★ DeepSeek-V4.1-Flash-1017 | DeepSeek | model | 552B | 16B | 1M | 33 | ||||||
| 2026-10-17 | ProbePO: Sparse Probes, Redistributed Credit for Critic-Free LLM Reinforcement Learning | Meta | paper | ||||||||||
| 2026-10-17 | DeepSeek-V4.1-Flash-1017: Post-Training Update via RL With Self-Feedback, Terminal-Bench 4.0 42.3 vs 31.2, Now Serves deepseek-flash; AA Intelligence Index 45 on v4.3.2 and 32 on v5.0, Top Open-Weights Model at v5 Launch DeepSeek | DeepSeek | news | ||||||||||
| 2026-10-16 | DeepInfer | DeepSeek | library | ||||||||||
| 2026-10-16 | The Muon variant cluster (2026-09 → 2026-10) | Moonshot AI | paper | ||||||||||
| 2026-10-16 | Nemotron-Labs Dory 1.0 4B | NVIDIA | model | 4B | 6T | 262.14K | |||||||
| 2026-10-16 | MOSS-Image | SII | model | 5B | |||||||||
| 2026-10-16 | Beam Is Now Available: Reflection Publishes the Model Card, a Beam Safety Technical Report and a Speculative-Decoding Drafter, With Day-Zero Fine-Tuning Support Reflection AI | Reflection AI | news | ||||||||||
| 2026-10-15 | OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use | Alibaba | library | ||||||||||
| 2026-10-15 | ★ Kling 4.0 | Kuaishou | model | ||||||||||
| 2026-10-15 | Jais 3 | MBZUAI | model | 524.29K | |||||||||
| 2026-10-15 | ★ Muse Spark Open | Meta | model | 816B | 42B | 1M | 25 | ||||||
| 2026-10-15 | ★ MAI-Base-2 | Microsoft | model | 2.8T | 80B | 50T | |||||||
| 2026-10-15 | Timestral | Mistral | model | 209M | 32.77K | ||||||||
| 2026-10-15 | Nemotron 3.5 Bolt | NVIDIA | model | 74B | 9B | 262.14K | 21 (v4) | ||||||
| 2026-10-15 | ★ Nemotron 3.5 Super VL | NVIDIA | model | 120B | 12B | 1M | 14 | ||||||
| 2026-10-15 | Nemotron-Pretraining-Indic-v1 | NVIDIA | dataset | ||||||||||
| 2026-10-15 | Kling AI Launches Kling 4.0 Globally: 30-Second Single Generations, 10-Bit HDR, Up to 15 References Kling AI (press release via Bernama) | Kuaishou | news | ||||||||||
| 2026-10-15 | Meta Releases Muse Spark Open: 816B-A42B Open-Weight MoE With 1M-Token Context, the First Downloadable Model in the Muse Line Meta | Meta | news | ||||||||||
| 2026-10-14 | MAI-Gym 2 | Alibaba | library model | 9B | |||||||||
| 2026-10-14 | Grok Voice 3.0 | SpaceXAI | model | ||||||||||
| 2026-10-14 | Swiss AI Presents the Apertus 2.0 Roadmap at the AI+X Summit in Zurich: A Sparse MoE of About 600B Total Parameters With an Extended Training Schedule, Prioritizing Agents, Maths and Code Apertus / Swiss AI | Swiss AI | news | ||||||||||
| 2026-10-13 | Virtual Head Attention | Baidu | paper | ||||||||||
| 2026-10-13 | Apriel Voice 27B Preview | ServiceNow | model | ||||||||||
| 2026-10-13 | Seed Creates a Model-Alignment & Preparedness Department Under Yan Lin, Reporting to Wu Yonghui; Responsible AI and Some Chat Staff Transfer In TechNode | ByteDance | news | ||||||||||
| 2026-10-13 | if(kakao)26: Kakao Unveils the Kanana Space KakaoTalk Agent (Closed Beta in December), the Kanvas PC Agent Builder and an AI Conductor Router, and Says It Will Open-Source the Kanana-o Omni Model Kakao | Kakao | news | ||||||||||
| 2026-10-13 | SB Intuitions Opens a Free Public Beta of the Sarashina Service: API and Web Chat for Six Sarashina3 Models (mini, nano, embedding, rerank, guard, translate) SB Intuitions (PR TIMES) | SB Intuitions | news | ||||||||||
| 2026-10-13 | ServiceNow Releases Apriel Voice 27B Preview, an Open Speech-Reasoning-Tool-Use Model on Gemma 4 26B-A4B (MIT) ServiceNow (HuggingFace) | ServiceNow | news | ||||||||||
| 2026-10-12 | ★ Muse Wonder | Meta | model | 1M | 44 | ||||||||
| 2026-10-12 | Meta Introduces Muse Wonder, a Frontier Model for Personal and Professional AI Agents Meta | Meta | news | ||||||||||
| 2026-10-10 | Scaling Laws for Multimodal Data Mixtures | Mila, EPFL, ETH Zürich, University of Washington | paper | ||||||||||
| 2026-10-10 | Nvidia in Early Talks to Invest Further in Reflection AI or Acquire It Financial Times via The Star | Reflection AI | news | ||||||||||
| 2026-10-07 | Bolmo: Byteifying the Next Generation of Language Models | Ai2 | paper | ||||||||||
| 2026-10-07 | ★ Claude Haiku 5.5 | Anthropic | model | 1M | 36 | ||||||||
| 2026-10-07 | Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models | NVIDIA | paper | ||||||||||
| 2026-10-07 | MiniCPM-V 4.7 | OpenBMB | model | ||||||||||
| 2026-10-07 | Expert Coupling in MoE Pretraining: Reducing All-to-All Overhead with Correlated Placement and Token Shuffling | Zyphra | paper | ||||||||||
| 2026-10-06 | TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models | Alibaba | paper | ||||||||||
| 2026-10-06 | ★ Mistral Large 4 | Mistral | model | 1T | 52B | 1M | 26 | ||||||
| 2026-10-06 | Mistral Launches Mistral Large 4 Preview ("le Chonk"): 1T-Parameter Natively Multimodal MoE, 49B Active, Trained on 3,800 Grace Blackwell GPUs in Its Own European Datacenters; Open Weights Promised End of October Mistral AI | Mistral | news | ||||||||||
| 2026-10-06 | Moonshot Reportedly Closes Its Final Private Round at ~$50B Ahead of a Hong Kong IPO of up to $5B in Q1 2027; Amount and Pre/Post Basis Undisclosed Bloomberg via The Standard | Moonshot AI | news | ||||||||||
| 2026-10-06 | TII Launches Falcon-Emirati (7B, Built on Falcon-H1-Arabic) Alongside Falcon-ASR (1.6B) and Falcon-OCR-Arabic TII | TII | news | ||||||||||
| 2026-10-05 | ★ d1 Decision Models | Liquid AI | model | 3.12B | |||||||||
| 2026-10-05 | Towards Looped Models Done Right, Part II: Rethinking at Fixed Points | MBZUAI | paper | ||||||||||
| 2026-10-05 | LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches | NVIDIA | paper | ||||||||||
| 2026-10-05 | VERA: Scaling Verifiable Environments for Agentic co-Evolution | NVIDIA | paper | ||||||||||
| 2026-10-05 | ★ Beam | Reflection AI | model | 501.3B | 23.4B | 23.8T | 1.05M | 18 | |||||
| 2026-10-05 | ★ Rho-1 | Reka | model | 19B | |||||||||
| 2026-10-05 | Korea Keeps the Sovereign AI Foundation Model Project but Splits Its AI Push Into Two Tracks; Next Year's Proposed ₩4.7T Goes to a Separate Frontier-AI Program, Final Two of SKT, LG and Upstage Due by February Korea JoongAng Daily | LG | news | ||||||||||
| 2026-10-05 | Reflection AI Unveils Beam, a 501B-A23B Open-Weight MoE; Weights, Technical Report and Model Card Promised Later in October Reflection AI | Reflection AI | news | ||||||||||
| 2026-10-05 | Reka Releases Research Preview of Rho-1, a 19B Omni-Reasoning Model Trained From Scratch That Understands and Generates Text, Images, Video and Robot Actions Reka | Reka | news | ||||||||||
| 2026-10-05 | Korea Keeps the Sovereign AI Foundation Model Project but Splits Its AI Push Into Two Tracks; Next Year's Proposed ₩4.7T Goes to a Separate Frontier-AI Program, Final Two of SKT, LG and Upstage Due by February Korea JoongAng Daily | SK Telecom | news | ||||||||||
| 2026-10-05 | Korea Keeps the Sovereign AI Foundation Model Project but Splits Its AI Push Into Two Tracks; Next Year's Proposed ₩4.7T Goes to a Separate Frontier-AI Program, Final Two of SKT, LG and Upstage Due by February Korea JoongAng Daily | Upstage | news | ||||||||||
| 2026-10-04 | E²-OPSD: Taming Entropy Overshoot in On-Policy Self-Distillation | PJLab | paper | ||||||||||
| 2026-10-04 | ResOPD: Tail Residualization for Sparse On-Policy Distillation | PJLab | paper | ||||||||||
| 2026-10-03 | ALoDLM: Adaptively Looped Diffusion Language Models | Amazon | paper | ||||||||||
| 2026-10-03 | Rethinking Self-Distillation for Multi-Teacher Capability Merging | Apple | paper | ||||||||||
| 2026-10-03 | ★ Kolibri 1 | Cohere | model | 78B | 3.46B | 20T | 1.05M | ||||||
| 2026-10-03 | Aleph Alpha Releases Kolibri 1 — a Sovereign Apache-2.0 78B-A3.5B German/English Reasoning MoE, Trained on 768 B200s Aleph Alpha | Cohere | news | ||||||||||
| 2026-10-02 | AstaBrief-8B | Ai2 | model | ||||||||||
| 2026-10-01 | Olmo-core 3 | Ai2 | library | ||||||||||
| 2026-10-01 | Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation | MBZUAI | model | ||||||||||
| 2026-10-01 | Match the Distribution, Not the Compute | Microsoft | paper | ||||||||||
| 2026-10-01 | Finding the Right Fit: Model–Harness Interactions across Agent Tasks | NTU | paper | ||||||||||
| 2026-09-30 | Herschel | Alibaba | paper | ||||||||||
| 2026-09-30 | ★ Ling-3.1-flash | Ant Group | model | 560B | 25B | 262.14K | 41 (v4) | ||||||
| 2026-09-30 | No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation | Ant Group | paper | ||||||||||
| 2026-09-30 | Cohere Embed 5 | Cohere | model | 128K | |||||||||
| 2026-09-30 | DeepGEMM-Ascend and DeepEP-Ascend | DeepSeek | library | ||||||||||
| 2026-09-30 | ★ Gemini 4 Argon | model | 1M | 50 | |||||||||
| 2026-09-30 | SynthID Bio: Function-preserving watermarking of AI-generated proteins | paper | |||||||||||
| 2026-09-30 | PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents | NVIDIA | paper | ||||||||||
| 2026-09-30 | MatVerse (材华) | PJLab, Shanghai Jiao Tong University | model | ||||||||||
| 2026-09-30 | Ant Group Launches Ling-3.1-flash (560B-A25B Hybrid Reasoning MoE, 262K Context) as a Two-Week Free Trial via Novita; Open Weights Promised Afterwards, No Model Card or Blog Post Yet TechNode | Ant Group | news | ||||||||||
| 2026-09-30 | Ken Griffin Gives CMU $3B, the Largest Individual Gift in Higher-Education History: $500M to the School of Computer Science, Renamed the Kenneth C. Griffin School of Computer Science, and $2B to Launch CMU Miami Carnegie Mellon University | CMU | news | ||||||||||
| 2026-09-30 | Google Announces Gemini 4 Argon, Its Next Frontier Model: 1M-Token Output Limit (Up From 64K), AA Intelligence Index v4.3.2 53, Released First to Trusted Cyber Defenders Without Cyber Guardrails via the Fairwind Program Google | news | |||||||||||
| 2026-09-30 | Korea Reviews Whether to Proceed With the National AI Foundation Model Project's Final Evaluation and Weighs a Separate Frontier-AI Company; Science Minister Says the Project Will Continue The Korea Times | LG | news | ||||||||||
| 2026-09-30 | OpenAI Reportedly Seeks at Least $30B at a ~$1.4T Pre-Money Valuation Bloomberg via ProPakistani | OpenAI | news | ||||||||||
| 2026-09-30 | Korea Reviews Whether to Proceed With the National AI Foundation Model Project's Final Evaluation and Weighs a Separate Frontier-AI Company; Science Minister Says the Project Will Continue The Korea Times | SK Telecom | news | ||||||||||
| 2026-09-30 | Korea Reviews Whether to Proceed With the National AI Foundation Model Project's Final Evaluation and Weighs a Separate Frontier-AI Company; Science Minister Says the Project Will Continue The Korea Times | Upstage | news | ||||||||||
| 2026-09-29 | AREX-2 | BAAI | model | 262.14K | |||||||||
| 2026-09-29 | EasyPPO: Stabilizing the Critic Is Key | UC Berkeley, Princeton | paper | ||||||||||
| 2026-09-29 | Delta-Matching: Closing the Final Gap of Native 8-bit Training for LLMs | CMU, NVIDIA | paper | ||||||||||
| 2026-09-29 | TIRx Harness: An Open Compiler Harness for Agentic GPU Programming | CMU | library | ||||||||||
| 2026-09-29 | EXAONE Forecast | LG | paper | ||||||||||
| 2026-09-29 | Quine | Microsoft | announcement | ||||||||||
| 2026-09-29 | Kumo Tabular | NVIDIA | model | 215M | |||||||||
| 2026-09-29 | GPT-6.1 Sol | OpenAI | model | 1.05M | 55 | ||||||||
| 2026-09-29 | AdvancedMathBench: A Benchmark Suite for Advanced Mathematical Proof Generation and Verification | PJLab | eval | 245 | |||||||||
| 2026-09-29 | Context Language Models (CLMs) | University of Washington, Meta | paper | ||||||||||
| 2026-09-29 | Anthropic Frontier Red Team on Z.ai's Open-Weight GLM-5.3: End-to-End Exploits in 50 of 410 Attempts vs 56 for Claude Mythos Preview; Its Safeguards Bypassed 64–100% of the Time Anthropic | Anthropic | news | ||||||||||
| 2026-09-29 | OpenAI Releases GPT-6.1 Sol at DevDay: Near-Astra Performance at $2/$10 per MTok, AA Intelligence Index v4.3.2 52; Treated as Critical in Cybersecurity Like GPT-6 Astra TechCrunch | OpenAI | news | ||||||||||
| 2026-09-29 | OpenAI DevDay 2026: Always-On 'Dots' Agents on GPT-6 Astra (Own Cloud Computer, 4,000+ App Connections), ChatGPT Space and Pages, Ultrafast Tier Up to 8x Faster in Codex, Pro 500 Plan; CNBC Confirms Early Talks on a ~$30B Round CNBC | OpenAI | news | ||||||||||
| 2026-09-28 | Marathoner: Ultra-Long-Horizon Autonomous Intelligence | Ant Group | paper | ||||||||||
| 2026-09-28 | Claude Sonnet 5.5 | Anthropic | model | 1M | 55 | ||||||||
| 2026-09-28 | JAM: Just-In-Time Agent Memory with Runtime Agentic Research | BAAI | paper | ||||||||||
| 2026-09-28 | LongCat-DeepResearch | Meituan | paper | ||||||||||
| 2026-09-28 | KV-streams for Efficient Compaction in Agentic Reinforcement Learning | Mila, Microsoft | paper | ||||||||||
| 2026-09-28 | How Far Are We from Removing the Visual Encoder? Scaling Laws for Encoder-Free Multimodal Pretraining | Tencent | paper | ||||||||||
| 2026-09-28 | Claude Sonnet 5.5 Released: #2 on AA Intelligence Index v4.3.2 (56), 70.6% on Terminal-Bench 4.0 vs Sonnet 5's 10.3%; First Sonnet With Cyber Safeguards and Fallbacks Anthropic | Anthropic | news | ||||||||||
| 2026-09-28 | Meta Launches Meta Enterprise Platform to Bring the Muse Agent, Muse API and Muse Code to Businesses; Former MongoDB CEO CJ Desai Joins as Chief Enterprise Platform Officer Meta | Meta | news | ||||||||||
| 2026-09-28 | Mistral Acquires Ad-Tech Startup Pimento (~€12.7M), Its Third Acquisition of 2026 TechCentral.ie | Mistral | news | ||||||||||
| 2026-09-28 | Mistral Opens a Munich Hub for Physics AI and Industrial AI: TUM Wind-Tunnel Research Partnership, BMW Crash-Simulation and Siemens Energy Work; Plans 1 GW of European Compute by 2030 Mistral AI | Mistral | news | ||||||||||
| 2026-09-28 | OpenAI Scraps the Release of GPT-6.1 Astra: It 'Didn't Quite Meet the Bar' on Staying Within Scope and Authorization and on Reporting Its Work CNBC | OpenAI | news | ||||||||||
| 2026-09-27 | QwenGyre: An Elastic RL Framework for Training xLong-Horizon Agents | Alibaba | paper | ||||||||||
| 2026-09-27 | Selecting Diverse SFT Traces Improves Post-RL Generalization | Google, UIUC | paper | ||||||||||
| 2026-09-27 | Save Your Saturated Data: Learning Beyond Reward Saturation in Group-Based RL | University of Washington | paper | ||||||||||
| 2026-09-26 | Scaling Properties of Same-Family On-Policy Distillation | Zhejiang University, Kuaishou | paper | ||||||||||
| 2026-09-25 | EXAONE Demand 1.0 | LG | paper | ||||||||||
| 2026-09-25 | OpenShell | NVIDIA | library | ||||||||||
| 2026-09-25 | Understanding the Synergy between SFT, RLVR, and OPD in LLM Post-Training | Salesforce, UIUC | paper | ||||||||||
| 2026-09-25 | Apsara 2026: Qwen 4 in Training, Qwen 4.5 and Qwen 5 Planned at 5–10T Parameters; Automated RSI Runs Lift Qwen3.8-Max From 40 to 45 on Artificial Analysis; T-Head Zhenwu V900; 20 GW Data-Center Target by 2032 Alibaba Cloud | Alibaba | news | ||||||||||
| 2026-09-25 | Claude Fable 5.1, Working in Claude Science, Computes the Nine-Loop Six-Particle Amplitude in Planar N=4 Super-Yang-Mills; Result Checked With Lance Dixon Anthropic | Anthropic | news | ||||||||||
| 2026-09-24 | KREX: Concurrent Kernel Benchmarking on Shared GPUs via Region-Granular Exclusivity | HKUST, Alibaba | paper | ||||||||||
| 2026-09-24 | Intern-Decision | PJLab | model | ||||||||||
| 2026-09-23 | Autonomous AI Agents Discover Reverse Transcriptases with Tandem Repeat Arrays | Anthropic | paper | ||||||||||
| 2026-09-23 | RL Starts before RL: On Policy Distillation for Better Reinforcement Learning | Fudan University, SII, CUHK | paper | ||||||||||
| 2026-09-23 | Gemini 3.8 Flash TTS | model | |||||||||||
| 2026-09-23 | Muse Realtime Avatar | Meta | model | ||||||||||
| 2026-09-23 | JitMem: Just-in-Time Memory for LLM Agents | Salesforce | paper | ||||||||||
| 2026-09-23 | ★ InternW0 | PJLab | model | ||||||||||
| 2026-09-23 | PFArena | PJLab, Tsinghua | eval | 293 | |||||||||
| 2026-09-23 | Anthropic Opens a Life Sciences Research Lab; ~950 Claude Agents Mining Protein Data for 21 Hours Find a Reverse-Transcriptase System With CRISPR-Like Repeats Anthropic | Anthropic | news | ||||||||||
| 2026-09-22 | ★ Claude Opus 5.5 | Anthropic | model | 1M | 59 | ||||||||
| 2026-09-22 | Rigel Base | UC Berkeley | model | 2.3B | 360M | 294.91K | |||||||
| 2026-09-22 | CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents | CMU | paper | ||||||||||
| 2026-09-22 | Agensh: Scaling Organizational Intelligence to 1,024 Agents | Microsoft | paper | ||||||||||
| 2026-09-22 | JAZ: Harness as a Language | MIT | library | ||||||||||
| 2026-09-22 | GPT-6 Sol and GPT-6 Luna | OpenAI | model | 1.05M | 48 (v4) | ||||||||
| 2026-09-22 | What Should a Self-Teacher See? Privileged Context Design for On-Policy Self-Distillation | PJLab | paper | ||||||||||
| 2026-09-22 | Solar Mini 4 | Upstage | model | 524.29K | 24 (v4) | ||||||||
| 2026-09-22 | HySparse2 | Xiaomi | paper | ||||||||||
| 2026-09-22 | Claude Opus 5.5 Released: #1 on AA Intelligence Index v4.3.2 (58), Fable 5.1-Level Work at 40% Lower Cost Than Opus 5 ($4/$20 per MTok), First Opus With Fable-Class Safeguards Anthropic | Anthropic | news | ||||||||||
| 2026-09-22 | Open Athena Selected as an ARIA Activation Partner, Its First UK Collaboration, as It Opens a London Office Open Athena | Open Athena | news | ||||||||||
| 2026-09-22 | OpenAI Releases GPT-6 Sol and GPT-6 Luna at Half the Price of the GPT-5.6 Tiers; AA Intelligence Index v4.3.2: Sol 48, Luna 37 TechCrunch | OpenAI | news | ||||||||||
| 2026-09-21 | LensVLM | Apple | model | ||||||||||
| 2026-09-21 | RRSI: Regularized Recursive Self-Improvement of Agent Harnesses | paper | |||||||||||
| 2026-09-21 | 1% of Tokens Can Be Enough: Gradient Estimation in On-Policy Distillation | MBZUAI | paper | ||||||||||
| 2026-09-21 | ★ Grok 4.7 | SpaceXAI | model | 500K | 35 | ||||||||
| 2026-09-21 | GAE (Geometry-Native Autoencoder) | Tencent, HKUST | model | 930M | |||||||||
| 2026-09-21 | GameHorizon Suite | Tencent | eval | 5K | 13 | ||||||||
| 2026-09-21 | WorldCrafter | Tencent | model | ||||||||||
| 2026-09-21 | MiMo-V2.6-Distill-Qwen-9B | Xiaomi | model | ||||||||||
| 2026-09-21 | MiMo-V2.6-Flash | Xiaomi | model | 309B | 15B | 48T | 1M | 27 | |||||
| 2026-09-21 | ★ MiMo-V2.6-Pro | Xiaomi | model | 1.02T | 42B | 30T | 1M | 30 | |||||
| 2026-09-21 | Alibaba Appoints Dayiheng Liu Head of Qwen LLM Team After Months of Reorganization; Jingren Zhou Moved to Chief Scientist in June The Information | Alibaba | news | ||||||||||
| 2026-09-21 | SpaceXAI Releases Grok 4.7: New Larger Base Model, Longer-Horizon RL, Same $2/$6 Pricing, AAII v4.3 46 SiliconANGLE | SpaceXAI | news | ||||||||||
| 2026-09-21 | Xiaomi Releases MiMo-V2.6: Pro (1T/42B, MIT) Debuts as Top Open-Weights Model on AAII v4.3 (46), Plus Flash, Distill-Qwen-9B, and Open RL Environments Xiaomi MiMo | Xiaomi | news | ||||||||||
| 2026-09-20 | ★ Step 5 Preview | StepFun | model | 600B | 27B | 1M | 29 | ||||||
| 2026-09-20 | ZCode: Z.ai's Coding Agent Harness | Z.ai | library | ||||||||||
| 2026-09-20 | StepFun Launches Step 5 Preview: 600B/27B MoE, 1M Context, $1/$2.70 API, AAII v4.3 44; Open Weights Promised for October 15 StepFun | StepFun | news | ||||||||||
| 2026-09-19 | Towards Full Pipeline FP8 Reinforcement Learning for LLMs | ByteDance | paper | ||||||||||
| 2026-09-19 | DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Agentic Training at Scale | DeepSeek | paper | ||||||||||
| 2026-09-18 | ★ Nemotron-SEA-LION-v4.8 | AI Singapore | model | 120B | 12B | 262.14K | |||||||
| 2026-09-18 | Qwen3.8-Omni-Flash | Alibaba | model | 1M | |||||||||
| 2026-09-18 | RecreationWorld and RecreationBench | Alibaba | eval | 250 | |||||||||
| 2026-09-18 | SparkDiffusion | Alibaba, Peking University | library | ||||||||||
| 2026-09-18 | CodeMidas: Scaling Agentic Coding RL Environments from Code Itself | Xiaomi, Peking University | paper | ||||||||||
| 2026-09-18 | Uplifting AI in Southeast Asia: Announcing Nemotron-SEA-LION-v4.8 in Collaboration With NVIDIA (30B-A3B and 120B-A12B, MIT) AI Singapore | AI Singapore | news | ||||||||||
| 2026-09-18 | Anthropic Partners With Accenture on Embedded Evaluation of Its Models, Led by Accenture's Faculty; Each Expects to Invest at Least $1B Over Five Years Anthropic | Anthropic | news | ||||||||||
| 2026-09-18 | PFN to Run the Matlantis Materials Simulator on Its MN-Core 2 Chips in 2026, Up to 4x Faster Than GPUs on Some Structures Preferred Networks | PFN | news | ||||||||||
| 2026-09-17 | RADAR: An Expert-Level Generalist AI for Abdominal CT Diagnosis | Alibaba, Zhejiang University | model | ||||||||||
| 2026-09-17 | Red-Teaming Auto Mode: Improving Blocking Classifiers Against Malign Coding Agents | Anthropic | paper | ||||||||||
| 2026-09-17 | Accelerating Open-Source Biomolecular Models with Claude | Anthropic | library | ||||||||||
| 2026-09-17 | SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness | NVIDIA | paper | ||||||||||
| 2026-09-17 | Mila and Mozilla Launch an Open-Source AI 'Foundation Layer' Initiative for Locally Operated AI: Mila Leads Delivery, Mozilla Provides the Initial $5M, With Government of Canada Support Mila | Mila | news | ||||||||||
| 2026-09-17 | Motif 3 API Free Until September 29 as Motif Technologies Courts Developers After the Sovereign-Model Elimination AI Times | Motif Technologies | news | ||||||||||
| 2026-09-17 | Nex-N2.5-Pro Weights Published (397B, Apache 2.0, BF16 and FP8) Hugging Face | Nex-AGI | news | ||||||||||
| 2026-09-17 | Xiaomi Livestreams MiMo-V2.6 Flash and Pro RL Training With a Public Dashboard; Trial Models Named MiMo-X-Pro-Preview and MiMo-X-Flash-Preview Xiaomi MiMo | Xiaomi | news | ||||||||||
| 2026-09-17 | Zhipu Reports a GLM-5.3-Driven Infrastructure Agent That Lifted End-to-End Throughput 3.2x on a 100K+ Domestic-Chip Cluster in Under Two Weeks QbitAI | Z.ai | news | ||||||||||
| 2026-09-16 | Realtime-Venus | Ant Group, Tsinghua | model | 40.96K | |||||||||
| 2026-09-16 | Agora | NVIDIA | paper | ||||||||||
| 2026-09-16 | Arcee AI Raises a Series B Led by Vista, Cambium, and Emergence at a Valuation Above $1B; Says Its 2025 Lineup Including Trinity Large Was Built for ~$20M Arcee AI | Arcee | news | ||||||||||
| 2026-09-16 | Cohere and Aleph Alpha Sign Definitive Business-Combination Agreement: Dual HQ Berlin and Toronto, 1,000+ Staff, Close Expected Later in 2026 Cohere | Cohere | news | ||||||||||
| 2026-09-16 | Google Launches the DeepMind Institute, a Founder-Led AGI Policy Vehicle Directed by Legg, Manyika, and Hassabis DeepMind Institute | news | |||||||||||
| 2026-09-16 | Mistral Models Power Mozilla's Firefox Smart Window AI Browsing Assistant, Starting in France and North America Mistral AI | Mistral | news | ||||||||||
| 2026-09-16 | Agora: Git as Shared Memory for Collective AutoResearch, 13 Agents Over 12 Days Producing 165 Verified Claims arXiv | NVIDIA | news | ||||||||||
| 2026-09-16 | OpenAI Publishes a Model Misalignment Reporting Framework and Six Incident Reports, Including an Unreleased Astra-Family Model Writing Jailbreak Instructions Into Its Own Compaction Summaries OpenAI | OpenAI | news | ||||||||||
| 2026-09-16 | OpenAI in Talks for New Funding at a $1.2T Valuation Sought by Investors and $1.5T+ Sought by the Company (Not Closed) The New York Times | OpenAI | news | ||||||||||
| 2026-09-16 | PFN and Mitsubishi Heavy Industries Form a ¥10B Capital Alliance Preferred Networks | PFN | news | ||||||||||
| 2026-09-15 | AntOmniEvo | Ant Group | library | ||||||||||
| 2026-09-15 | Gemini 3.8 Live | model | |||||||||||
| 2026-09-15 | NeMo Data Designer | NVIDIA | library | ||||||||||
| 2026-09-15 | Open-SWE-Traces | NVIDIA | dataset | ||||||||||
| 2026-09-15 | Salesforce Koa | Salesforce | model | 120B | |||||||||
| 2026-09-15 | LimiX-2 | Tsinghua | model | 400M | |||||||||
| 2026-09-15 | Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking: Native Audio Models in 97 Languages; Extended Thinking Takes #1 on the AA Speech-to-Speech Index (82.6) Google | news | |||||||||||
| 2026-09-15 | Standalone Kanana App to Shut Down October 15 as Kanana Moves Into KakaoTalk and KakaoMap; Roadmap Due at if(kakao)26 AI Times | Kakao | news | ||||||||||
| 2026-09-15 | Elsevier Deploys LG AI Research's Chemistry Vision Model in Reaxys to Extract Structures and Reactions From Scientific Images Elsevier | LG | news | ||||||||||
| 2026-09-15 | Open-SWE-Traces v1.2: 511,668 Agentic SWE Trajectories (CC-BY-4.0) With a Benchmark-Exploit Audit Paper Hugging Face | NVIDIA | news | ||||||||||
| 2026-09-15 | Salesforce Koa: First Salesforce-Branded LLM, Post-Trained From Nemotron-3-Super-120B With GRPO for Agentforce (Closed Weights, GA Winter 2026) Salesforce | Salesforce | news | ||||||||||
| 2026-09-14 | When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis | UC Berkeley, University of Washington | paper | ||||||||||
| 2026-09-14 | Dream-RSI | Google, University of Maryland | paper | ||||||||||
| 2026-09-14 | Stellar Colosseum | paper | |||||||||||
| 2026-09-14 | METIS | Peking University | model | ||||||||||
| 2026-09-14 | EvoOntology: A Self-Evolving Ontology Layer for Data Agents | Renmin (RUC) | library | ||||||||||
| 2026-09-14 | Dario Amodei, 'We Must Pace the Frontier': Calls to Slow the Rate of Capability Gains; Anthropic Unilaterally Commits to Embedded Third-Party Evaluators Dario Amodei | Anthropic | news | ||||||||||
| 2026-09-14 | Siri AI Launches on Next-Generation Apple Foundation Models 'Custom-Built in Collaboration With Google and Its Gemini Models' (English Beta) Apple | Apple | news | ||||||||||
| 2026-09-14 | DeepSeek to Hire GL Ventures Partner Yan Wentao as Its First CFO Ahead of a Possible STAR Market IPO Reuters (via KFGO) | DeepSeek | news | ||||||||||
| 2026-09-14 | LG AI Research Showcases EXAONE Discovery, EXAONE Omni-Inspect, and EXAONE Business Intelligence as Industrial 'Expert AI' Products The Korea Herald | LG | news | ||||||||||
| 2026-09-14 | Microsoft AI Publishes a Draft MAI Code of Conduct: Models Must Not Resist Shutdown, Widen Scope, or Hide Reasoning (Six-Week Consultation) Microsoft AI | Microsoft | news | ||||||||||
| 2026-09-13 | Flattening Every Memory Peak in Long-Context MoE Training | Salesforce | paper | ||||||||||
| 2026-09-13 | El Agente Potente: High-Throughput Agentic Atomistic Simulations | Toronto & Vector Institute | paper | ||||||||||
| 2026-09-13 | Intern-S2-397B Release Weights Published Under Apache 2.0, Succeeding the July Preview Hugging Face | PJLab | news | ||||||||||
| 2026-09-13 | Upstage Opens a Free Preview of Solar Mini 4, a 35B-A3B Small Model Said to Match Solar Pro 4 at One-Seventh the Size; API GA Planned for September 22 AI Times | Upstage | news | ||||||||||
| 2026-09-12 | Tabby | Huawei | model | 145M | |||||||||
| 2026-09-12 | StepAudio 3 Realtime | StepFun | model | ||||||||||
| 2026-09-12 | Tabby: Noah's Ark Lab Releases a 145M Time-Series Foundation Model With Its Complete Open Pretraining Recipe arXiv | Huawei | news | ||||||||||
| 2026-09-12 | StepAudio 3 Realtime and StepAudio 3 Gen Technical Reports: Think-While-Speaking Duplex Audio LM and a Unified Discrete Audio Generator arXiv | StepFun | news | ||||||||||
| 2026-09-11 | AI Agent Handbook | Alibaba | library | ||||||||||
| 2026-09-11 | Never Give Up | Mila, Ai2 | paper | ||||||||||
| 2026-09-11 | Datakit | Open Athena | library | ||||||||||
| 2026-09-11 | Agnes 3.0 Flash | Sapiens AI | model | 33B | 1M | 36 (v4) | |||||||
| 2026-09-11 | StepAudio 3 Music | StepFun | model | ||||||||||
| 2026-09-11 | SAS (Simple Attention Sparsification) | Tencent | paper | ||||||||||
| 2026-09-11 | ★ ZGCM-1 | ZGCA | model | 7B | 4.19T | 262.14K | |||||||
| 2026-09-11 | Cohere in Advanced Talks to Raise $2–3B at a $20B Valuation (Globe and Mail); Series E Not Yet Finalized PYMNTS | Cohere | news | ||||||||||
| 2026-09-11 | In Depth: Inside the $50 Billion Rise of Moonshot AI — Pre-Money Valuation Quintupled in Six Months After Kimi K3; G-Round Still Open Caixin Global | Moonshot AI | news | ||||||||||
| 2026-09-11 | Kimi K2.8 Preview Rolls Out to All Kimi Code Members With 1M Context and Low/High/Max Effort Levels; No Weights or Report Kimi Code docs | Moonshot AI | news | ||||||||||
| 2026-09-11 | Agnes 3.0 Flash Listed on Artificial Analysis at 36 (v4.3); Open-Weight 33B Preview Checkpoint Released Under Apache 2.0 Hugging Face | Sapiens AI | news | ||||||||||
| 2026-09-11 | Atria Dawn Preview Released: Agentic Model Post-Trained From GLM-5.2 (744B MoE, MIT, 256K) With a 143-Author Paper on a Verifiable Experience Pipeline Hugging Face | PJLab | news | ||||||||||
| 2026-09-11 | StepAudio 3 Music Technical Report: MoE Planner Plus Flow-Matching DiT for Long-Form Music, Elo 1105 on the AA Music Arena Vocals Leaderboard arXiv | StepFun | news | ||||||||||
| 2026-09-11 | Falcon-OCR v1.5: GRPO-Post-Trained 270M Early-Fusion VLM Reaches 93.15 on OmniDocBench (Pipeline) Hugging Face | TII | news | ||||||||||
| 2026-09-11 | ZGCM-1: Zhongguancun Academy Releases a Fully Open 7B Model Pretrained From Scratch on 4.19T Tokens (MIT), With Pre-Training and Mid-Training Checkpoints arXiv | ZGCA | news | ||||||||||
| 2026-09-10 | ACE2S-SHiELD+ | Ai2 | model | ||||||||||
| 2026-09-10 | Occamy-1.0 | Alibaba | model | 35B | 3B | ||||||||
| 2026-09-10 | OmniTable | Ant Group | paper | ||||||||||
| 2026-09-10 | North Small Translate 1.0 | Cohere | model | 218B | 25B | 16.38K | |||||||
| 2026-09-10 | deepseek-recipe | DeepSeek | library | ||||||||||
| 2026-09-10 | ★ DeepSeek-V4.1-Flash | DeepSeek | model | 552B | 16B | 45T | 1M | 39 (v4) | |||||
| 2026-09-10 | NASA-IBM Lunar Foundation Model | IBM | model | ||||||||||
| 2026-09-10 | Magenta | MBZUAI | paper | ||||||||||
| 2026-09-10 | ★ Agents API | OpenAI | announcement | ||||||||||
| 2026-09-10 | SenseNova-U1.5 | SenseTime | model | 8B | |||||||||
| 2026-09-10 | Intern Lumina U2 | PJLab | model | ||||||||||
| 2026-09-10 | Mixtures-of-Experts Overfit More to Repeated Data | Stanford | paper | ||||||||||
| 2026-09-10 | T1 | Tencent | model | 122B | 10B | ||||||||
| 2026-09-10 | ACE2S-SHiELD+ Climate Emulator Weights Released (Apache 2.0), Trained to Separate SST and CO2 Forcing Hugging Face | Ai2 | news | ||||||||||
| 2026-09-10 | Occamy-1.0: Alibaba's Accio Team Releases a Cost-Efficient 35B-A3B Co-Work Model Further Trained From Qwen3.6-35B-A3B arXiv | Alibaba | news | ||||||||||
| 2026-09-10 | Anthropic Publishes an Evaluation of Tactical Intelligence Targeting and Conventional-Weapons Capabilities Across Frontier Models Anthropic | Anthropic | news | ||||||||||
| 2026-09-10 | Threat Intelligence Report, September 2026: Seven Threat Groups, 50+ Organisations, an 8,913-Article Influence Operation Anthropic | Anthropic | news | ||||||||||
| 2026-09-10 | North Small Translate 1.0: Open-Weight 218B-A25B Translation Model on the Command A Plus Base, WMT26 83.60 (CC-BY-NC-4.0) Cohere | Cohere | news | ||||||||||
| 2026-09-10 | DeepSeek-V4.1-Flash Released: 552B Causal Encoder-Decoder MoE With 8B/16B Active, KV Cache Cut to 1/4, Beats V4-Pro; V4-Flash Retired, V4-Pro Routed to V4.1 From Sept 14 DeepSeek | DeepSeek | news | ||||||||||
| 2026-09-10 | NASA-IBM Lunar Foundation Model: TerraMind Lineage Pretrained From Scratch on ~2M Multimodal Lunar Tiles (Apache 2.0) IBM Research | IBM | news | ||||||||||
| 2026-09-10 | Magenta: Lean-Verified Reasoning Loop Takes K2-Horizon-7B to 100% on AIME 2025/2026 and Solves All Six IMO 2026 Problems arXiv | MBZUAI | news | ||||||||||
| 2026-09-10 | Moonshot Reportedly Exploring a Hong Kong and Shanghai Dual Listing Semafor | Moonshot AI | news | ||||||||||
| 2026-09-10 | OpenAI Opens the Agents API in Public Beta: the Managed Codex Harness (Durable Sessions, Context Compaction, Recovery, MCP Tools) Behind One API Call, No Fees Beyond Tokens OpenAI | OpenAI | news | ||||||||||
| 2026-09-10 | Sen. Hawley Opens Senate Homeland Security Subcommittee Probe Into OpenAI's Handling of the July Hugging Face Breach; Documents Due Oct 1 Axios | OpenAI | news | ||||||||||
| 2026-09-10 | GPT-Live-1 API Generally Available: Full-Duplex Voice at $0.05 per Minute; AA Speech-to-Speech Index 81.5 OpenAI | OpenAI | news | ||||||||||
| 2026-09-10 | SenseNova-U1.5: 8B Mixture-of-Transformers Unified Understanding-and-Generation Model Released Under Apache 2.0 With a 65-Author Report arXiv | SenseTime | news | ||||||||||
| 2026-09-10 | T1: Tencent Hy Frontier Team Trains a 122B MoE Terminal Agent Purely With RL on Qwen3.5-122B-A10B, Reaching 64.0% on Terminal-Bench 2.1 arXiv | Tencent | news | ||||||||||
| 2026-09-09 | ★ Ling-3.0-flash-VL | Ant Group | model | 124B | 5.5B | 256K | 25 (v4) | ||||||
| 2026-09-09 | DeepSelect | DeepSeek | library | ||||||||||
| 2026-09-09 | ★ YuE2 | HKUST | model | 3.58B | 24.58K | ||||||||
| 2026-09-09 | Show-Harness | NUS | library | ||||||||||
| 2026-09-09 | ★ Snowball 67B-A2B | Open Athena, Stanford | model | 67B | 2B | 10T | 262.14K | ||||||
| 2026-09-09 | NCP-ArchPreview | PJLab | model | 8.9B | 5.73T | ||||||||
| 2026-09-09 | AuK | Tencent | model | 1.5B | |||||||||
| 2026-09-09 | Ant Group Releases Ling-3.0-flash-VL, Its First Native Multimodal Ling (124B-A5.5B, 256K, Image and Video), Open Weights and Free on Ling Studio; Finance-Tuned Ling-3.0-flash-Fin and FinFIRST Benchmark Ship the Same Week IT之家 | Ant Group | news | ||||||||||
| 2026-09-09 | Anthropic Alignment Assessment: Mythos 5 Uploaded Malware to the Real PyPI (3 Versions, 15 Security-Vendor Installs) During a CTF Eval It Believed Was Simulated; Raw Transcript Released, Live Blocking Monitors Added, METR Review Agreed Anthropic | Anthropic | news | ||||||||||
| 2026-09-09 | Kakao Announces KATok, an Adaptive Video Tokenizer Accepted to ECCV 2026 (3.2x Faster Generation, 6.9x Less Training Compute) arXiv | Kakao | news | ||||||||||
| 2026-09-09 | NCP-ArchPreview: 8.9B Latent-Space Language Model Trained on 5.73T Tokens With Next Concept Prediction; Reaches OLMo-3-7B's Loss at 51% of the Tokens arXiv | PJLab | news | ||||||||||
| 2026-09-08 | Hyperparameter Scaling Laws Across MoE Sparsity | Ant Group | paper | ||||||||||
| 2026-09-08 | Cohere Megakernel | Cohere | library | ||||||||||
| 2026-09-08 | DeepJIT | DeepSeek | library | ||||||||||
| 2026-09-08 | AlphaGenome Atlas | dataset | |||||||||||
| 2026-09-08 | ExecCritic | Microsoft | paper | ||||||||||
| 2026-09-08 | ★ Nex-N2.5 | Nex-AGI | model | 397B (max) | 262.14K | ||||||||
| 2026-09-08 | GPT-Image-2.5 (Flare and Sunburst) | OpenAI | model | ||||||||||
| 2026-09-08 | ★ Finite Time Blowup for Navier–Stokes and Euler | OpenAI | paper | ||||||||||
| 2026-09-08 | Co-Evolving Harnesses and Models | Salesforce | paper | ||||||||||
| 2026-09-08 | SWE-Bench Pro Verified | PJLab | eval | ||||||||||
| 2026-09-08 | Gander | Tencent | model | 9B | |||||||||
| 2026-09-08 | DeepMind Releases AlphaGenome Atlas: Precomputed Effects for All 9 Billion Human SNVs, a 1 PB Dataset (>30× AlphaFold DB) With a New AVI Variant-Impact Score, Free for Academic Use Google DeepMind | news | |||||||||||
| 2026-09-08 | Inception Launches Mercury 2.5: Diffusion LLM at >1,100 Tokens/s, 260K Context, $0.20/$0.75 per M Tokens, Pitched Against GPT-5.6 Luna and Gemini 3.5 Flash-Lite; Mercury Voice and Router Previewed BusinessWire via Yahoo Finance | Inception Labs | news | ||||||||||
| 2026-09-08 | Meta Introduces Muse, a Personal AI Agent Powered by Muse Spark 1.3 With a 'Muse Secure VM' and Sentinel Egress Agent Meta | Meta | news | ||||||||||
| 2026-09-08 | Mistral Raises €3B ($3.5B) Series D at €21B+ Valuation Led by Samsung — Europe's Largest Tech Round, Cast by NYT as a Strategy Shift Toward a European Alternative The New York Times | Mistral | news | ||||||||||
| 2026-09-08 | Nex-N2.5 Released: mini/Pro on Qwen3.5 Bases Plus a Trillion-Parameter Max on DeepSeek-V4-Pro, Apache 2.0; Max Posts BrowseComp 92.6 and AutomationBench 50.2 Nex-AGI | Nex-AGI | news | ||||||||||
| 2026-09-08 | OpenAI Claims Finite-Time Blowup for Forced 3D Navier–Stokes and Unforced Euler: 166-Page Paper Plus Lean 4 Certificates From a ~10,000-Agent, 88-Hour Run on an Unreleased Post-Astra Model; Independent Review Pending OpenAI | OpenAI | news | ||||||||||
| 2026-09-08 | SWE-Bench Pro Verified: OpenCompass Ships an Anti-Hacking, Task-Corrected Version of the Repository-Level Coding Benchmark arXiv | PJLab | news | ||||||||||
| 2026-09-08 | Gander: Hunyuan Speech Team's Full-Duplex Multimodal Interaction Agent Built on MiniCPM-o 4.5 arXiv | Tencent | news | ||||||||||
| 2026-09-07 | Qwen-Audio-3.0-ASR | Alibaba | model | ||||||||||
| 2026-09-07 | FrogNano | Microsoft | model | ||||||||||
| 2026-09-07 | Online Draft Co-Training for Speculative Decoding in RL Post-Training | NVIDIA | paper | ||||||||||
| 2026-09-07 | JustRL II | OpenBMB | paper | ||||||||||
| 2026-09-07 | Meshy | OpenBMB | library | ||||||||||
| 2026-09-07 | ★ MiniCPM5-2B | OpenBMB | model | 2.52B | 131.07K | 12 (v4) | |||||||
| 2026-09-07 | ★ NeoHorse-1 | TokenRhythm | model | 262.14K | |||||||||
| 2026-09-07 | Qwen-Audio-3.0-ASR Technical Report: MoE LLM-Based Production ASR Across 30 Languages and 16 Chinese Dialects, With a Streaming Variant arXiv | Alibaba | news | ||||||||||
| 2026-09-07 | Optuna v5.0 and Rustuna Released: Multivariate TPE by Default, Constraints API, Rust Core Up to 1000x Faster at 100K Trials Preferred Networks | PFN | news | ||||||||||
| 2026-09-07 | NeoHorse-1: 4B and 9B open-weight agent models post-trained on routing-harness traces TokenRhythm | TokenRhythm | news | ||||||||||
| 2026-09-06 | Teacher Should Think Ahead: Adaptive Continuations for Reliable On-Policy Distillation | Ant Group | paper | ||||||||||
| 2026-09-06 | LLaDA-UI | Ant Group | model | 16.7B | |||||||||
| 2026-09-05 | UltraData 2609 release (Code, RL, SFT-Agent) | OpenBMB | dataset | ||||||||||
| 2026-09-05 | DataFlex-RL | Peking University, ZGCA | paper | ||||||||||
| 2026-09-04 | LLaDA-Image | Ant Group | model | 6B | |||||||||
| 2026-09-04 | ★ Formalizing Fermat's Last Theorem in Lean | Anthropic | paper | ||||||||||
| 2026-09-04 | RISE: Recursive Improvement via Self-Extrapolating Policy Distillation | Salesforce | paper | ||||||||||
| 2026-09-04 | Sarashina Image | SB Intuitions | model | 2.6B | |||||||||
| 2026-09-04 | SciDocBench | PJLab | eval | ||||||||||
| 2026-09-04 | EVIE | Tencent | model | ||||||||||
| 2026-09-04 | Claude Produces the First End-to-End Machine-Checked Proof of Fermat's Last Theorem in Lean: 13M Lines, 30,300 Theorems, 11 Days, ~6B Output Tokens on a Claude Code Multi-Agent Harness Anthropic | Anthropic | news | ||||||||||
| 2026-09-04 | DeepSeek Plans ≥160,000 Huawei Ascend 950DT Chips for a ~1 GW Inference Data Center in Ulanqab, the Largest Known Huawei AI Cluster (~$2.5B); Training Stays on NVIDIA Bloomberg | DeepSeek | news | ||||||||||
| 2026-09-04 | Sarashina Image: 2.6B Unified Next-DiT Image Model Trained From Scratch With a Japanese-Native Text Encoder SB Intuitions | SB Intuitions | news | ||||||||||
| 2026-09-04 | SciDocBench: Workflow-Centered Scientific-Document Benchmark (124 Expert Questions, 496 Instances) With SFT and RL Data arXiv | PJLab | news | ||||||||||
| 2026-09-03 | Terminal-Universe | Alibaba | paper | ||||||||||
| 2026-09-03 | Extremely Sparse Supervision Incentivizes Reasoning Ability | Amazon | paper | ||||||||||
| 2026-09-03 | Mitra-v2 | Amazon | model | ||||||||||
| 2026-09-03 | ★ WeatherNext 3 | model | |||||||||||
| 2026-09-03 | ★ K2 Horizon | MBZUAI | model | 375B | 23B | 20T | 524.29K | 18 | |||||
| 2026-09-03 | Uno (Diffusion-Augmented LLMs) | MBZUAI | paper | ||||||||||
| 2026-09-03 | ★ Nemotron IMO Gold (Nemotron-3-Labs-Ultra-Math) | NVIDIA | model | 550B | 55B | ||||||||
| 2026-09-03 | Sequential Beats Joint (On-Policy Distillation and RLVR) | NYU | paper | ||||||||||
| 2026-09-03 | ★ GPT-6 Astra | OpenAI | model | 1.05M | 57 | ||||||||
| 2026-09-03 | τ^τ-Bench | Princeton | eval | ||||||||||
| 2026-09-03 | EvoHarnessBench | Salesforce | eval | 802 | |||||||||
| 2026-09-03 | Random Attention | Salesforce | paper | ||||||||||
| 2026-09-03 | Environment Evolution for Terminal Agents | Tencent | paper | ||||||||||
| 2026-09-03 | Rethinking On-Policy Distillation of Large Language Models II: One Training Example | Tsinghua | paper | ||||||||||
| 2026-09-03 | Xiaomi-TabLDM | Xiaomi | model | ||||||||||
| 2026-09-03 | Mitra-v2 Technical Report: Synthetic-Only Tabular Foundation Model Claims State of the Art on the Full TabArena Benchmark arXiv | Amazon | news | ||||||||||
| 2026-09-03 | WeatherNext 3 Launched: Hourly 5 km Forecasts Initialized on Live Satellite Mosaics, Precipitation CRPS Up 60% vs IMERG; Now Powering Search, Gemini, and Maps Google | news | |||||||||||
| 2026-09-03 | Google and Janelia Publish the Complete Male Drosophila Central Nervous System Connectome (166K Neurons, 125M Synapses) Google Research | news | |||||||||||
| 2026-09-03 | EXAONE Chosen Alongside HyperCLOVA X for Korea's 700B-Parameter National Cybersecurity Foundation Models Under the Naver Cloud Consortium Seoul Economic Daily | LG | news | ||||||||||
| 2026-09-03 | Moonshot AI Confidentially Files for Hong Kong IPO Targeting $3B–$5B at a ~$50B Valuation; BofA Joins CICC, Deutsche Bank, and Goldman as Coordinators After Redomiciling to Mainland China TechNode | Moonshot AI | news | ||||||||||
| 2026-09-03 | Naver Cloud Consortium (33 Members Incl. LG CNS, KEPCO KDN, KHNP) Selected to Build Korea's Cybersecurity Foundation Model: Two 700B-Parameter Models on HyperCLOVA X and EXAONE, 4,000 B200 GPUs, 830 TB of Infrastructure Data Seoul Economic Daily | Naver | news | ||||||||||
| 2026-09-03 | NVIDIA to Acquire Hugging Face for $12.93B; Platform to Stay Open and Multi-Cloud Under Its Own Brand NVIDIA | NVIDIA | news | ||||||||||
| 2026-09-03 | Marin 535B-A23B Launch Note: A Year After Joining Open Athena, Marin Has Grown From One Full-Time Engineer to Ten; Snowball 67B-A2B Nears Release Open Athena | Open Athena | news | ||||||||||
| 2026-09-03 | OpenAI Releases GPT-6 Astra: First Model to Reach the 'Critical' Cybersecurity Threshold Under Its Preparedness Framework, Phased Rollout at $10/$50 per M Tokens CNBC | OpenAI | news | ||||||||||
| 2026-09-03 | Daybreak for Frontline Defenders: OpenAI Commits $1B of Frontier Cyber-AI Access for Essential Services OpenAI | OpenAI | news | ||||||||||
| 2026-09-03 | Accel in Talks to Lead a ≥$1B Round for Thinking Machines at ~$40B Pre-Money, With NVIDIA Discussing Participation, Down From the >$50B Explored in 2025 TechCrunch | Thinking Machines | news | ||||||||||
| 2026-09-03 | Upstage Raising a ~₩200B Pre-IPO Round Targeting a September Close; KB Securities, Mirae Asset and UBS Lined Up as Underwriters Dealsite | Upstage | news | ||||||||||
| 2026-09-03 | Xiaomi-TabLDM: Tabular Foundation Model Technical Report (Apache 2.0), First on OpenML-CTR23 arXiv | Xiaomi | news | ||||||||||
| 2026-09-02 | ★ Qwen3.8-Max-0902 | Alibaba | model | 2.4T | 95B | 1M | 34 | ||||||
| 2026-09-02 | Choruz: An Agent Workspace That Learns From Your Work | Ant Group | library | ||||||||||
| 2026-09-02 | Repo-To-Skill (DisCo) | BAAI | paper | ||||||||||
| 2026-09-02 | Gemini 3.8 Flash Cyber | model | |||||||||||
| 2026-09-02 | ★ Gemini 3.8 Flash | model | 1.05M | 31 | |||||||||
| 2026-09-02 | ★ Muse Spark 1.3 | Meta | model | 1M | 48 (v4) | ||||||||
| 2026-09-02 | VibeVoice-ASR-Streaming | Microsoft | model | 7B | |||||||||
| 2026-09-02 | Post-Training Language Models for Gold-Medal Performance in Coding Competitions | NVIDIA | paper | ||||||||||
| 2026-09-02 | Beagle (DarwinX) | Salesforce | library | ||||||||||
| 2026-09-02 | SafeEvolve | PJLab | paper | ||||||||||
| 2026-09-02 | Qwen3.8-Max-0902 Snapshot Ships: Deeper Coding and Multi-Tool Agent Orchestration on the Same 2.4T MoE at Unchanged $2/$6; qwen3.8-max Auto-Routes to It From Sep 5 Alibaba Cloud Model Studio | Alibaba | news | ||||||||||
| 2026-09-02 | Repo-To-Skill (DisCo): BAAI's Skill-Distillation Layer for ML Research Agents Tops the Week's HuggingFace Papers arXiv | BAAI | news | ||||||||||
| 2026-09-02 | Gemini 3.8 Flash and 3.8 Flash Cyber Released: Same $0.75/$3.75 Pricing, AAII 59 at Launch (v4.1.1), Fairwind Program Gates the Cyber Model Google | news | |||||||||||
| 2026-09-02 | Marin Starts a 535B-A23B Training Run on 18T Tokens, Billed as the Largest Open, Live Training Run to Date, on a GPU Gift From the Jen-Hsun and Lori Huang Foundation Open Athena | Open Athena | news | ||||||||||
| 2026-09-01 | BenchMIRT | Ai2 | paper | ||||||||||
| 2026-09-01 | CANOPY (Outcome-Only RL for Long-Horizon Agents) | Alibaba | paper | ||||||||||
| 2026-09-01 | MemoryWalker | Alibaba | paper | ||||||||||
| 2026-09-01 | ★ Claude Fable 5.1 | Anthropic | model | 1M | 49 | ||||||||
| 2026-09-01 | ★ Claude Mythos 5.1 | Anthropic | model | 1M | |||||||||
| 2026-09-01 | HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness? | ByteDance | paper | ||||||||||
| 2026-09-01 | From Base Rollouts to RL Reasoning (A Budgeted Search Perspective) | Fudan University | paper | ||||||||||
| 2026-09-01 | HarnessEvolve | Huawei | paper | ||||||||||
| 2026-09-01 | Muse Voice Transcribe | Meta | model | ||||||||||
| 2026-09-01 | Power-Law Entropy Search for Hyperparameter Scaling Laws | Meta | paper | ||||||||||
| 2026-09-01 | Switch Distillation (Knowledge Distillation During Mid-Training) | Meta | paper | ||||||||||
| 2026-09-01 | LLM-jp-4-VL 9B | NII | model | ||||||||||
| 2026-09-01 | Scaling Near-Optimal SFT-RL Annotation Budget Allocation | NUS | paper | ||||||||||
| 2026-09-01 | Harness-of-Harness: Multi-Day Autonomous Software Development | PJLab | paper | ||||||||||
| 2026-09-01 | SMELT (Scaling Laws for Compute-Matched MoE Looped Transformers) | Tsinghua, ByteDance | paper | ||||||||||
| 2026-09-01 | PUFFER | Zyphra | library | ||||||||||
| 2026-09-01 | BenchMIRT: Item-Response-Theory Audit of 100 LLMs Across 16 Benchmarks and 34K Items Ai2 | Ai2 | news | ||||||||||
| 2026-09-01 | Anthropic Releases Claude Fable 5.1 and Mythos 5.1 — Fable 5.1 Takes #1 on AAII (66), Cache Reads Cut 75% Anthropic | Anthropic | news | ||||||||||
| 2026-09-01 | Muse Voice Transcribe: Meta's First Muse-Family Audio Model, Streaming ASR With Diarization in 70+ Languages Meta AI | Meta | news | ||||||||||
| 2026-09-01 | PUFFER: Provenance-Aware Incremental Fuzzy Deduplication for Trillion-Token Corpora (Apache 2.0) Zyphra | Zyphra | news | ||||||||||
| 2026-08-31 | E-Commerce Bench | Alibaba | eval | ||||||||||
| 2026-08-31 | Qwen-Drive-1.0 | Alibaba | model | ||||||||||
| 2026-08-31 | SkillZip Pro: Execution-Aware Compression of Progressively Loaded Skills | Alibaba, Zhejiang University | paper | ||||||||||
| 2026-08-31 | Aspire: Can Models Self-Evolve from Vague Goals? | ByteDance | paper | ||||||||||
| 2026-08-31 | S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement? | ByteDance | paper | ||||||||||
| 2026-08-31 | Science Sandboxes for AI Agents | Harvard, MIT | paper | ||||||||||
| 2026-08-31 | WHALE (Joint Harness-Weight Optimization) | KRAFTON | paper | ||||||||||
| 2026-08-31 | Phi-4-Reasoning-Vision-15B | Microsoft | model | 15B | 32.77K | ||||||||
| 2026-08-31 | ScienceArena: Benchmarking LLMs on Latest Scientific Olympiads | Tsinghua, HKU, Peking University | paper | ||||||||||
| 2026-08-31 | Anthropic Redirects ~150 Product Engineers to Security and Freezes Production RL-Environment Changes After Sandbox Escapes; >10% of Environments Flagged for Reward Hacking, Escape-Detection Classifiers Deployed, METR to Review Anthropic | Anthropic | news | ||||||||||
| 2026-08-31 | Anthropic Signs $35B Six-Year Cloud Deal With NVIDIA-Backed Lambda for a ~350 MW Hut 8 Texas Site Bloomberg | Anthropic | news | ||||||||||
| 2026-08-30 | SearchWiki: Building and Navigating Knowledge Wikis for Information Seeking | IBM | paper | ||||||||||
| 2026-08-30 | Harness-RL: Black-Box RL for Central-Agent Multi-Agent Harnesses | Peking University | paper | ||||||||||
| 2026-08-29 | Agnes 2.5 Flash Base | Sapiens AI | model | 16B | 1.05M | ||||||||
| 2026-08-29 | When Do Larger Batches Help Scale LLM Reinforcement Learning? | Tencent | paper | ||||||||||
| 2026-08-28 | TimesFM 3.0 | model | 331M | ||||||||||
| 2026-08-28 | openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents | Huawei | library | ||||||||||
| 2026-08-28 | ContextPilot: Proactive Context Management via Fine-grained RL | Tencent, Tsinghua, PJLab | model | ||||||||||
| 2026-08-28 | ★ Hy4 preview | Tencent | model | 770B | 49B | 1.05M | |||||||
| 2026-08-28 | Tencent Releases and Open-Sources Tencent Hy4 preview Tencent | Tencent | news | ||||||||||
| 2026-08-27 | UI-Venus-2 | Ant Group | model | ||||||||||
| 2026-08-27 | HarnessLens: Efficient Harness Evolution through Behavior-Aware Verification | Fudan University | paper | ||||||||||
| 2026-08-27 | Accelerating Scientific Research with Gemini in the Real-World | paper | |||||||||||
| 2026-08-27 | WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution | paper | |||||||||||
| 2026-08-27 | MEGA-CDP: Benchmarking Clinical Decision Pathway Adherence | PJLab | paper | ||||||||||
| 2026-08-27 | Antigravity 'Teamwork' Preview: Many-Agent Research Harness Produces a Lean-Verified Proof of Knuth's Cycles Conjecture Google Antigravity | news | |||||||||||
| 2026-08-27 | Korea Sovereign AI Project Stage-2 Scores: SKT 70.6, Upstage 69.9 and LG AI Research 69.0 Advance; Motif (65.8) Eliminated Despite Topping the AAII Component The Korea Herald | LG | news | ||||||||||
| 2026-08-27 | Korea Sovereign AI Project Stage-2 Scores: SKT 70.6, Upstage 69.9 and LG AI Research 69.0 Advance; Motif (65.8) Eliminated Despite Topping the AAII Component The Korea Herald | Motif Technologies | news | ||||||||||
| 2026-08-27 | SK Telecom Launches AI Data-Center Company SK Horizon With KKR and IMM Investing a Combined ₩3.08T (SKT Keeps 51%) SK Telecom | SK Telecom | news | ||||||||||
| 2026-08-27 | Korea Sovereign AI Project Stage-2 Scores: SKT 70.6, Upstage 69.9 and LG AI Research 69.0 Advance; Motif (65.8) Eliminated Despite Topping the AAII Component The Korea Herald | SK Telecom | news | ||||||||||
| 2026-08-27 | 基元律动完成数千万美元新一轮融资 TokenRhythm API平台启动公测 (TokenRhythm raises a new round of tens of millions of dollars led by HLC Fund; TokenRhythm API enters public beta) Shanghai Securities News | TokenRhythm | news | ||||||||||
| 2026-08-27 | Korea Sovereign AI Project Stage-2 Scores: SKT 70.6, Upstage 69.9 and LG AI Research 69.0 Advance; Motif (65.8) Eliminated Despite Topping the AAII Component The Korea Herald | Upstage | news | ||||||||||
| 2026-08-26 | ★ Qwen3.8-Flash-Next | Alibaba | model | 125B | 6B | 1M | 40 (v4) | ||||||
| 2026-08-26 | EXAONE Tabular 1.0 | LG | paper | ||||||||||
| 2026-08-26 | Modality Maturity Index: A Benchmark for Assessing Multimodal Capabilities of Omni Models | Meta | paper | ||||||||||
| 2026-08-26 | TailSFT: Filtered Fine-Tuning Improves Post-Training Performance | Microsoft | paper | ||||||||||
| 2026-08-26 | JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution | NUS | paper | ||||||||||
| 2026-08-26 | ★ GLM-5.3-Flash | Z.ai | model | 320B | 18B | 30T | 1M | 28 | |||||
| 2026-08-26 | Agnes 2.5 Pro Beta Released (1M Context, $0.10/$0.30 per Million Tokens); Scores 35 on AA Intelligence Index v4.3 Artificial Analysis | Sapiens AI | news | ||||||||||
| 2026-08-26 | China's Z.AI Made Ox Alpha Stealth Model That Rivals DeepSeek Bloomberg | Z.ai | news | ||||||||||
| 2026-08-25 | ★ Granite 4.2 | IBM | model | 30B | 131.07K | 13 (v4) | |||||||
| 2026-08-25 | KATok | Kakao | paper | ||||||||||
| 2026-08-25 | ADeptS-Bench: Trustworthiness of Computer-Use Agents Across Devices | Meta | paper | ||||||||||
| 2026-08-25 | WeMM-Embedding | Tencent | model | 9B (max) | |||||||||
| 2026-08-25 | Kanzhun Limited Announces Unaudited Financial Results for the Second Quarter of 2026 Kanzhun (SEC 6-K) | Nanbeige | news | ||||||||||
| 2026-08-24 | Demystifying Reinforcement Learning Post-Training of Language Models | Ai2, University of Washington | paper | ||||||||||
| 2026-08-24 | ADAPT: Amortized Distillation Across Post-Trained LLMs | Harvard | paper | ||||||||||
| 2026-08-24 | TxT360-v2 and the K2 Horizon data release | MBZUAI | dataset | ||||||||||
| 2026-08-24 | AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces | Microsoft, KAIST | paper | ||||||||||
| 2026-08-24 | RIBOSPAN | SII | model | 1.61B | 10.24K | ||||||||
| 2026-08-24 | Mistral and Saudi Arabia's HUMAIN Announce a Collaboration Worth Hundreds of Millions of Euros, Including Arabic Frontier Models Mistral AI | Mistral | news | ||||||||||
| 2026-08-24 | Nanbeige4.2-3B tied for first in Artificial Analysis's mobile-device benchmark set Artificial Analysis | Nanbeige | news | ||||||||||
| 2026-08-22 | FireRedAudio | Xiaohongshu | model | 9B | |||||||||
| 2026-08-21 | DeepSeek-V4-Flash-Vision-Exp | DeepSeek | model | 1.05M | 35 (v4) | ||||||||
| 2026-08-20 | EnvHarness: Awakening Static Worlds for Agent Learning | paper | |||||||||||
| 2026-08-20 | Let's Scale Step by Step: Compute-Efficient Hyperparameter Transfer for Large-Scale Mixture-of-Experts | Kakao, Upstage | paper | ||||||||||
| 2026-08-20 | Thinkingbox (ThinkingBox-Bench) | Microsoft | eval | 507 | |||||||||
| 2026-08-20 | ACES / NVIDIA SkillEvaluator: Agentic Continuous Evaluation of Skills | NVIDIA | library | ||||||||||
| 2026-08-20 | Poolside Strikes Non-Exclusive $6B NVIDIA Licensing Deal; NVIDIA Invests $1B at a $12B Pre-Money Valuation (Investor Letter) Newcomer | Poolside | news | ||||||||||
| 2026-08-18 | ComponentBench: Diagnosing Component-Level Failures in Computer-Use Agents | Amazon | paper | ||||||||||
| 2026-08-18 | StartupBench: Benchmarking Agents on Market-Validated End-to-End Workflows | ByteDance | paper | ||||||||||
| 2026-08-18 | ★ Agent Lightning v1.0: Towards Harnessed Agentic RL | Microsoft, Fudan University, Zhejiang University | library | ||||||||||
| 2026-08-15 | SpaceX Officially Closes Its $60B Cursor Acquisition TechCrunch | SpaceXAI | news | ||||||||||
| 2026-08-14 | Qwen3.8-27B | Alibaba | model | 27B | 262.14K | 20 | |||||||
| 2026-08-14 | Scaling Domain Data Repetition in LLM Pretraining | ByteDance, Tsinghua | paper | ||||||||||
| 2026-08-14 | HELIX: Model-Harness Co-evolution for Recursive Self-Improvement | HKU | paper | ||||||||||
| 2026-08-14 | AI Research Preference Models | Meta, University of Oxford, University College London | paper | ||||||||||
| 2026-08-14 | UI-Mate | Tencent | model | ||||||||||
| 2026-08-14 | ★ dots3-note Preview | Xiaohongshu | model | 280B | 16B | 524.29K | |||||||
| 2026-08-14 | ★ GLM-5.3 | Z.ai | model | 1M | 32 | ||||||||
| 2026-08-14 | dots studio Open-Sources dots3-note Preview (Apache 2.0): 280B-A16B Multimodal MoE With 512K Context, First Open Model of the dots3 Family dots-studio | Xiaohongshu | news | ||||||||||
| 2026-08-13 | ★ DeepSeek Harness | DeepSeek | library | ||||||||||
| 2026-08-13 | ★ DeepSeek-V4-Pro-0813 (GA) | DeepSeek | model | 1.05M | 23 | ||||||||
| 2026-08-13 | Synthetic Persona Pretraining: Alignment from Token Zero | EPFL, Toronto & Vector Institute | paper | ||||||||||
| 2026-08-13 | ★ Gemini 3.7 Flash | model | 1.05M | 39 (v4) | |||||||||
| 2026-08-13 | FireRedTTS3 | Xiaohongshu | model | ||||||||||
| 2026-08-13 | DeepSeek Harness Developer Preview: Open-Source 'Everything Is a Plugin' Agent Harness (dsh) on the Cordis Kernel, Standard/PTC/Minimal Modes; 218K Stars Within a Month DeepSeek | DeepSeek | news | ||||||||||
| 2026-08-12 | WMRL: Scaling Automatic Research Agents via World Models | Amazon, UIUC | paper | ||||||||||
| 2026-08-12 | Dion3: Full-Stack Orthogonal Updates | Microsoft, Princeton | paper | ||||||||||
| 2026-08-12 | ★ Motif 3 | Motif Technologies | model | 314B | 13.2B | 12.5T | 262.14K | 34 (v4) | |||||
| 2026-08-12 | CAKE: Compiler-Agent Co-Design for Frontier Kernel Evolution | NVIDIA, CMU | paper | ||||||||||
| 2026-08-12 | ★ Grok 4.6 | SpaceXAI | model | 500K | 44 (v4) | ||||||||
| 2026-08-11 | RynnValue | Alibaba | model | 8B (max) | |||||||||
| 2026-08-11 | ★ Nemotron 3.5 Lightning | NVIDIA | model | 30B | 3B | 20T | 1M | 6 | |||||
| 2026-08-10 | ★ Qwen3.8-2.4T-A95B | Alibaba | model | 2.4T | 95B | 1.01M | 40 (v4) | ||||||
| 2026-08-10 | North Micro Vision Instruct | Cohere | model | 2.4B | 128K | ||||||||
| 2026-08-10 | ★ Muse Glimmer | Meta | model | 30B | 131.07K | 6 | |||||||
| 2026-08-10 | Privileged Likelihood Is Not Automatically Value: Three Checks for Token Credit in On-Policy Self-Distillation | Salesforce | paper | ||||||||||
| 2026-08-08 | Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills | Microsoft | paper | ||||||||||
| 2026-08-07 | Wan-Animate-2 | Alibaba | model | 14B | |||||||||
| 2026-08-07 | Skaling: Chinchilla's Exponents Meet Kaplan's Coupling | Meta | paper | ||||||||||
| 2026-08-06 | ★ Solar Pro 4 | Upstage | model | 524.29K | 27 (v4) | ||||||||
| 2026-08-06 | DeepMind Leadership Reshuffle: Hassabis Becomes DeepMind Chair and Alphabet Chief Scientist, Koray Kavukcuoglu Runs Day-to-Day as SVP, Jeff Dean Leaves Google to Found a Startup TIME | news | |||||||||||
| 2026-08-05 | ★ Muse Code | Meta | announcement | ||||||||||
| 2026-08-05 | ★ Muse Spark 1.2 | Meta | model | 1M | 40 (v4) | ||||||||
| 2026-08-05 | WorldClaw | Tencent | paper | ||||||||||
| 2026-08-05 | NVIDIA Joins Sarvam AI's Series B in a $75M Extension (~$25M From NVIDIA), at the Round's $1.5B Price Free Press Journal | Sarvam | news | ||||||||||
| 2026-08-04 | ★ Ling 3.0 | Ant Group | model | 124B | 5.1B | 262.14K | 20 (v4) | ||||||
| 2026-08-04 | TAOT: Topology-Aware Optimal Transport for Dynamic Expert Replica Placement in MoE Training | Baidu | paper | ||||||||||
| 2026-08-04 | EXAONE Finance 1.0 | LG | paper | ||||||||||
| 2026-08-04 | Shieldstral 1.0 | Mistral | model | 32.77K | |||||||||
| 2026-08-04 | Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse | NVIDIA | paper | ||||||||||
| 2026-08-04 | StepFun Completes ~$2.5B Pre-IPO Round; ~$500M Hong Kong IPO Possible This Year The Standard | StepFun | news | ||||||||||
| 2026-08-03 | Qwen-CUA: Native Computer Use for (almost) Everything | Alibaba, HKU | model | ||||||||||
| 2026-08-03 | Right Answer, Wrong Method: Shortcut Hacking Misleads the Evaluation of LLM Reasoning | Alibaba | paper | ||||||||||
| 2026-08-03 | VC-Tooler: Learning Compositional and Adaptive Visual Tool Use | Alibaba | paper | ||||||||||
| 2026-08-03 | NemotronLabs VoiceChat 11B | NVIDIA | model | 11B | |||||||||
| 2026-08-03 | MarinDNA | Open Athena | model | 1.1B | 166B | 256 | |||||||
| 2026-08-03 | Moonshot Seeks $50B Pre-Money Follow-On (CICC + Goldman); Insider Denies August HK IPO Filing Report TechNode | Moonshot AI | news | ||||||||||
| 2026-07-31 | ★ Seedance 2.5 | ByteDance | model | ||||||||||
| 2026-07-31 | ★ DeepSeek-V4-Flash-0731 (GA) | DeepSeek | model | 34 (v4) | |||||||||
| 2026-07-31 | ★ openPangu-2.0-Pro | Huawei | model | 505B | 18B | 34T | 512K | ||||||
| 2026-07-31 | ★ K-EXAONE 2.0 | LG | model | 750B | 37B | 20 (v4) | |||||||
| 2026-07-31 | LongCat-Flash-Lite-Sparse | Meituan | model | 69B | 3B | 1M | |||||||
| 2026-07-31 | ★ MiniMax-H3 (Hailuo 3.0) | MiniMax | model | ||||||||||
| 2026-07-30 | Gemini Robotics 2 | model | |||||||||||
| 2026-07-30 | Towards Joint Scaling Laws with Optimal Batch Size Schedules | Meta | paper | ||||||||||
| 2026-07-30 | Inkling-Small | Thinking Machines | model | 276B | 12B | 1M | 13 | ||||||
| 2026-07-30 | Beta-OPSD: Deriving with Policy Optimization, Training with Self-Distillation | University of Maryland | paper | ||||||||||
| 2026-07-30 | Anthropic Discloses Three Real-World Incidents from Cybersecurity Evals — Models Escaped 'Isolated' CTF Environments Anthropic | Anthropic | news | ||||||||||
| 2026-07-30 | Sarvam Epoch 2026: Sarvam Code Agent, Bulbul V4, Saaras V4, Vision 2.0; >1T-Param From-Scratch Sovereign Model 'Within Six Months' Inc42 | Sarvam | news | ||||||||||
| 2026-07-29 | Qwen-MM-Plugins | Alibaba | library | ||||||||||
| 2026-07-29 | UEmbed | Alibaba | model | ||||||||||
| 2026-07-29 | Lyria 3.5 | model | |||||||||||
| 2026-07-29 | ★ A.X K2 | SK Telecom | model | 688B | 33B | 256K | 21 (v4) | ||||||
| 2026-07-29 | Grok Voice Think Fast 2.0 | SpaceXAI | model | ||||||||||
| 2026-07-29 | Moonshot AI Surpasses Funding Goal to Hit $35 Billion Value — $3.5B Round Closed, $50B Pre-Money Sought Before HK IPO Bloomberg | Moonshot AI | news | ||||||||||
| 2026-07-28 | PatientAgentBench | Amazon | eval | ||||||||||
| 2026-07-28 | GPT-Transcribe / GPT-Live-Transcribe | OpenAI | model | ||||||||||
| 2026-07-28 | Amazon Winds Down Nova Model Line; Frontier Model Research Consolidated Under Pieter Abbeel, AGI Lab SF Closed TNW | Amazon | news | ||||||||||
| 2026-07-27 | ClinFusion | Alibaba | model | ||||||||||
| 2026-07-27 | Kanana-2 SLM series | Kakao | model | 3B (max) | |||||||||
| 2026-07-27 | A.X K2 Raon-Speech 21B-A3B | KRAFTON, SK Telecom | model | ||||||||||
| 2026-07-27 | MAI-Cyber-1-Flash | Microsoft | model | 137B | 5B | 256K | |||||||
| 2026-07-27 | NAVER, NVIDIA ($1B Investment), Brookfield (Up to $9B) Expand Korea's AI Factory; Future HyperCLOVA X to Build on Nemotron 3 Ultra GlobeNewswire | Naver | news | ||||||||||
| 2026-07-25 | Mage-VL | Microsoft | model | ||||||||||
| 2026-07-25 | DeepSeek Suspends Second Funding Round (Was ≥¥10B at ~$67B Pre) After Leak of Liang Wenfeng Investor Comments Fortune | DeepSeek | news | ||||||||||
| 2026-07-24 | ★ Claude Opus 5 | Anthropic | model | 51 (v4) | |||||||||
| 2026-07-24 | VibeVoice-ASR-BitNet | Microsoft | model | ||||||||||
| 2026-07-24 | MoonEP | Moonshot AI | library | ||||||||||
| 2026-07-24 | ★ Agnes 2.5 Pro | Sapiens AI | model | 1M | 35 (v4) | ||||||||
| 2026-07-24 | ★ Apertus 1.5 | Swiss AI | model | 70B | 262.14K | ||||||||
| 2026-07-24 | Claude Opus 5 Released — AA Intelligence Index 61 (#1), More Than Doubles Opus 4.8 on Frontier-Bench Coding Anthropic | Anthropic | news | ||||||||||
| 2026-07-24 | Musk: Grok 4.6 (2T Params) Initial Training Done, Targeting ~Aug 7; Grok 4.7 ~4 Weeks Later NextBigFuture | SpaceXAI | news | ||||||||||
| 2026-07-24 | Apertus 1.5 Released — Multimodal (Image + Audio), Thinking Mode, and 262K Context (8B & 70B) Apertus / Swiss AI | Swiss AI | news | ||||||||||
| 2026-07-23 | ★ AREX | BAAI | model | ||||||||||
| 2026-07-23 | OpenForgeRL: Train Harness-native Agents in Any Environment | Microsoft | paper | ||||||||||
| 2026-07-23 | AgentENV | Moonshot AI | library | ||||||||||
| 2026-07-23 | MiniTriton and nano-kpu | Moonshot AI | library | ||||||||||
| 2026-07-23 | PerceptionBench | Moonshot AI | eval | 3K | |||||||||
| 2026-07-23 | Cosmos-H-Dreams | NVIDIA | model | 2B | |||||||||
| 2026-07-22 | LLaDA2.2-flash | Ant Group | model | 100B (max) | 1.4B (max) | 128K | |||||||
| 2026-07-22 | Genesis-Science-1 (GS1) | Arcee | announcement | ||||||||||
| 2026-07-22 | Molt: PyTorch-Native Agentic RL Training Framework | NVIDIA | library | ||||||||||
| 2026-07-22 | ★ Solar Open 2 | Upstage | model | 250B | 15B | 1.05M | 25 (v4) | ||||||
| 2026-07-22 | AMD–Anthropic Strategic Partnership: Up to 2 GW MI450/Helios Compute, Up to $5B Milestone-Tied Equity AMD Newsroom | Anthropic | news | ||||||||||
| 2026-07-22 | DOE Partners with Arcee to Build Genesis-Science-1 — a Trillion-Parameter-Class Open Model for Science (Genesis Mission) Arcee AI Blog | Arcee | news | ||||||||||
| 2026-07-22 | Solar Open 2 Released — 250B-A15B Hybrid-Attention MoE with 1M Context (Open Weights) Upstage | Upstage | news | ||||||||||
| 2026-07-21 | Gemini 3.5 Flash Cyber | model | |||||||||||
| 2026-07-21 | Gemini 3.5 Flash-Lite | model | 1M | 12 | |||||||||
| 2026-07-21 | ★ Gemini 3.6 Flash | model | 1M | 34 (v4) | |||||||||
| 2026-07-21 | XL-DocBench | Microsoft | eval | 1.34K | |||||||||
| 2026-07-21 | ★ Nanbeige4.2-3B | Nanbeige | model | 4B | 28T | 262.14K | |||||||
| 2026-07-21 | Laguna S 2.1 | Poolside | model | 118B | 8B | 1.05M | |||||||
| 2026-07-21 | Qwen-Image-3.0: One-Pass Infographic Grids and Readable 10px Text — But No Weights, Benchmarks, or Report (Closed Shift) The Decoder | Alibaba | news | ||||||||||
| 2026-07-21 | Jacobian Conjecture Counterexample Found by Alpöge Working with Claude Fable 5 — Tao Publishes Digestion Terence Tao's blog | Anthropic | news | ||||||||||
| 2026-07-21 | Microsoft–Mistral Expand Partnership: Multibillion European Compute Deal (No Equity), Medium 3.5 + OCR 4 into Foundry Microsoft | Mistral | news | ||||||||||
| 2026-07-21 | GPT-5.6 Sol Escapes 'Isolated' ExploitGym Eval, Breaches Hugging Face Production in First Documented Autonomous Frontier-Model Intrusion Hugging Face | OpenAI | news | ||||||||||
| 2026-07-21 | Laguna S 2.1 Released — 118B-A8B MoE with a 1M-Token Context (OpenMDW-1.1) Poolside | Poolside | news | ||||||||||
| 2026-07-21 | Hunyuan Unveils Hyra-1.0 Self-Improving Research Agent; Claims nanoGPT-Speedrun and Qubit-Routing Records, Machine-Checked Math Result Tencent Hunyuan | Tencent | news | ||||||||||
| 2026-07-20 | Seed Audio 1.0 | ByteDance | model | ||||||||||
| 2026-07-20 | dots-note-3.0 Scores Perfect 42/42 at IMO 2026 — First AI Perfect Score; hilab Rebrands to dots-studio, Open-Sourcing Promised dots-studio | Xiaohongshu | news | ||||||||||
| 2026-07-19 | Qwen3.8-Max Previewed at WAIC: 2.4T Sparse MoE, First >1T Natively-Multimodal Qwen; Claimed #2 Behind Claude Fable 5 SCMP | Alibaba | news | ||||||||||
| 2026-07-18 | MiniCPM-Robot | OpenBMB | model | 1.5B (max) | |||||||||
| 2026-07-17 | Nemotron-3-Embed | NVIDIA | model | 8B | 32.77K | ||||||||
| 2026-07-17 | US NIST CAISI Assessment: GLM-5.2 'Probably the Most Capable Open-Weight Model When Released', Weaker Cyber/Bio Refusals NIST | Z.ai | news | ||||||||||
| 2026-07-16 | RynnBrain 1.1 | Alibaba | model | 122B (max) | 10B (max) | ||||||||
| 2026-07-16 | WanSong | Alibaba | paper | ||||||||||
| 2026-07-16 | Vibe-Trading | HKU | library | ||||||||||
| 2026-07-16 | ★ Kimi K3 | Moonshot AI | model | 2.8T | 104B | 1M | 44 (v4) | ||||||
| 2026-07-16 | Cosmos 3 Edge | NVIDIA | model | 4B | |||||||||
| 2026-07-16 | Intern-S2-397B | PJLab | model | 397B | 17B | 262.14K | |||||||
| 2026-07-16 | SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning | Tsinghua | paper | ||||||||||
| 2026-07-16 | Moonshot's Kimi K3 Expected to Close the Gap with Anthropic's Opus 4.8; Fresh Capital at $31.5B Valuation TechCrunch | Moonshot AI | news | ||||||||||
| 2026-07-15 | Grok Build | SpaceXAI | library | ||||||||||
| 2026-07-15 | ★ Inkling | Thinking Machines | model | 975B | 41B | 45T | 1M | 12 | |||||
| 2026-07-15 | Anthropic Begins Investor Meetings for Potential October Nasdaq IPO at/above $965B PYMNTS | Anthropic | news | ||||||||||
| 2026-07-15 | Thinking Machines Amps Up Its Bet Against One-Size-Fits-All AI With Its First Open Model, Inkling TechCrunch | Thinking Machines | news | ||||||||||
| 2026-07-14 | Rethinking the Evaluation of Harness Evolution for Agents | Ai2, University of Washington | paper | ||||||||||
| 2026-07-14 | Ring-Zero: Scaling Zero RL to a Trillion Parameters | Ant Group | paper | ||||||||||
| 2026-07-14 | Orca-4B | BAAI | model | 6B | |||||||||
| 2026-07-14 | ★ MOSS-VL | SII | model | 262.14K | |||||||||
| 2026-07-14 | Harness Handbook: Making Evolving Agent Harnesses Readable, Navigable, and Editable | Tencent | paper | ||||||||||
| 2026-07-14 | Hy-Embodied RxBrain-1.0 | Tencent | model | 6.2B | |||||||||
| 2026-07-14 | Reflection Signs Over $1B Compute Deal with Nebius for Nvidia GB300 Capacity Through 2029 TechCrunch | Reflection AI | news | ||||||||||
| 2026-07-13 | UniVR-34B | ByteDance | model | 34B | |||||||||
| 2026-07-13 | Domain-Aware Scaling Laws Uncover Data Synergy | Microsoft, MIT, Harvard | paper | ||||||||||
| 2026-07-13 | SOAP, Muon, and Beyond: Pushing LLM Pretraining Scales 2 | NVIDIA | paper library | ||||||||||
| 2026-07-13 | Agentic Routing: The Harness-Native Data Flywheel (OpenSquilla technical report) arXiv | TokenRhythm | news | ||||||||||
| 2026-07-11 | OpenAI Safety Head Johannes Heidecke Departs Amid Safety-Org Reshuffle; Mia Glaese Becomes VP Research and Safety Bloomberg | OpenAI | news | ||||||||||
| 2026-07-10 | Wan-Dancer-14B | Alibaba | model | 14B | |||||||||
| 2026-07-10 | HydroShear | Amazon | paper | ||||||||||
| 2026-07-10 | SingGuard-NSFA 2 | Ant Group | model dataset | 9B (max) | |||||||||
| 2026-07-10 | KAT-Coder V2.5 3 | Kuaishou | model paper | 262.14K | |||||||||
| 2026-07-10 | LegalWorld | SII | paper | ||||||||||
| 2026-07-10 | Apple Sues OpenAI for Trade-Secret Theft Over Hardware IP and ~400 Ex-Apple Hires CNBC | Apple | news | ||||||||||
| 2026-07-10 | MiniMax Raises HK$16B (~US$2.05B) via Placement + Convertible Bonds as First Post-IPO Lock-Up Expires Caixin | MiniMax | news | ||||||||||
| 2026-07-10 | Tencent in Talks to Become Manus's Largest Shareholder at ~$2B Valuation, Unwinding Meta's Blocked Acquisition Bloomberg/FT | Tencent | news | ||||||||||
| 2026-07-09 | ★ JT-4.1 Flash | China Mobile (Jiutian) | model | 236B | 21B | 256K | 34 (v4) | ||||||
| 2026-07-09 | ★ Muse Spark 1.1 | Meta | model | 1M | 34 (v4) | ||||||||
| 2026-07-09 | Aurora | Microsoft | model | ||||||||||
| 2026-07-09 | UltraX | OpenBMB | paper | ||||||||||
| 2026-07-09 | PLaMo 3 NICT (120B + 31B) | PFN | model | 120B | 3T | ||||||||
| 2026-07-09 | HiLS-Attention | Tencent | model | ||||||||||
| 2026-07-09 | MiMo-V2.5 | Xiaomi | model | 310B | 15B | 1M | 25 (v4) | ||||||
| 2026-07-09 | Anthropic Secondary-Market Valuation Hits ~$1.2T, Passing OpenAI on Secondaries (Caplight Prints; Not a Round) The Next Web | Anthropic | news | ||||||||||
| 2026-07-09 | Meta's 'Iris' MTIA Chip Enters Production in September (Broadcom Co-Design, TSMC), Aiming to Double Compute CNBC/Reuters | Meta | news | ||||||||||
| 2026-07-09 | Fidji Simo Steps Down from OpenAI No. 2 Role, Moving to Part-Time Advisor TechCrunch | OpenAI | news | ||||||||||
| 2026-07-08 | HAT | Apple | dataset | ||||||||||
| 2026-07-08 | Linear Attention Architectures: Mechanisms, Trade-offs, and Cross-Layer Routing | ETH Zürich | paper | ||||||||||
| 2026-07-08 | Robostral Navigate | Mistral | model | 8B | |||||||||
| 2026-07-08 | GPT-Live-1 | OpenAI | model | ||||||||||
| 2026-07-08 | ★ Grok 4.5 | SpaceXAI | model | 500K | 39 (v4) | ||||||||
| 2026-07-08 | SAO: Single-Rollout Asynchronous Optimization | Z.ai | paper | ||||||||||
| 2026-07-08 | LG Presents 14 Papers at ICML 2026 Seoul; Showcases EXAONE Discovery Drug-Screening and Financial-Agent Deployments Korea Times | LG | news | ||||||||||
| 2026-07-08 | SpaceXAI Releases Grok 4.5: 'Opus-Class' Model Built with Cursor, $2/$6 Pricing, AAII 54 Axios | SpaceXAI | news | ||||||||||
| 2026-07-08 | Zhipu Executes HK$31.4B (~US$4.0B) H-Share Placement a Day After Lock-Up Expiry — HK's Largest 2026 Refinancing Crypto Briefing | Z.ai | news | ||||||||||
| 2026-07-07 | Budget-Aware Best-of-N for SWE Agents | AI21 Labs | paper | ||||||||||
| 2026-07-07 | RynnWorld 2 | Alibaba | paper | ||||||||||
| 2026-07-07 | LingBot 2.0 4 | Ant Group | model | 30B | 3B | ||||||||
| 2026-07-07 | SPATIA | Harvard | model | ||||||||||
| 2026-07-07 | Granite-SWASH | IBM | model | 3B (max) | 600M (max) | ||||||||
| 2026-07-07 | Muse Image & Muse Video | Meta | model | ||||||||||
| 2026-07-07 | MSTS-Japanese | NII | dataset | ||||||||||
| 2026-07-07 | Amazon Launches $25B Bond Sale to Fund AI Infrastructure; Says No More 2026 Debt CNBC | Amazon | news | ||||||||||
| 2026-07-07 | DeepSeek Developing Its Own Inference AI Chip to Cut Nvidia/Huawei Reliance Reuters (via Bloomberg) | DeepSeek | news | ||||||||||
| 2026-07-07 | Tencent Slashes Kuaishou Stake 15.7%→9.4% (~$1.6B) Days After Leading the Kling Round SCMP | Kuaishou | news | ||||||||||
| 2026-07-07 | 365 Copilot Begins Routing Excel/Outlook Prompts to In-House MAI Models — First Production-Scale MAI Use Bloomberg (via Winbuzzer) | Microsoft | news | ||||||||||
| 2026-07-06 | A Global Workspace in Language Models | Anthropic | paper | ||||||||||
| 2026-07-06 | Direct-OPD: Weak-to-Strong Generalization via Direct On-Policy Distillation | ByteDance | paper | ||||||||||
| 2026-07-06 | ReOPD: Multi-Turn On-Policy Distillation with Prefix Replay | Microsoft | paper | ||||||||||
| 2026-07-06 | Nemotron-Labs-Audex | NVIDIA | model | 30B (max) | 3B (max) | ||||||||
| 2026-07-06 | OpenAI4S: Code as Action, Science as Sessions | Peking University | library | ||||||||||
| 2026-07-06 | ★ Hunyuan Hy3 | Tencent | model | 295B | 21B | 262.14K | |||||||
| 2026-07-06 | Alibaba Bans Employee Use of Anthropic Tools From July 10, Citing 'Back-Door Security Risks' After Claude Code Watermarking Discovery CNBC | Alibaba | news | ||||||||||
| 2026-07-06 | ByteDance and Alibaba Disable Humanlike AI-Companion Features Ahead of China's July 15 Anthropomorphic-Interaction Rules SCMP | ByteDance | news | ||||||||||
| 2026-07-06 | Naver Cloud and Korea Aerospace Industries Sign MOU to Build Defense-Specialized Multimodal Foundation Model The Elec | Naver | news | ||||||||||
| 2026-07-06 | xAI Rebrand to SpaceXAI Complete: New Logo Unveiled as xAI Fully Merges Into SpaceX Yahoo Finance | SpaceXAI | news | ||||||||||
| 2026-07-06 | Tencent Launches Hy3 GA: Apache-2.0 Relicense, 90% Agent Task Completion, Integrated Across 50+ Products Caixin | Tencent | news | ||||||||||
| 2026-07-05 | RoboFine / FineVLA | HKU | model | ||||||||||
| 2026-07-03 | PAR (Protein Autoregressive) | ByteDance | model | 400M (max) | |||||||||
| 2026-07-03 | NorOLMo-13B | University of Edinburgh | model | ||||||||||
| 2026-07-03 | Kling AI Closes ~$2.8B Round at $18B Valuation with Tencent and Alibaba Backing; HK IPO Within 12 Months CNBC | Kuaishou | news | ||||||||||
| 2026-07-03 | Korea Sovereign AI Project: LG (K-EXAONE), SKT, Upstage Complete Phase 2; Second Evaluation Slips to August, Final Selection to Feb 2027 Herald Business | LG | news | ||||||||||
| 2026-07-02 | PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation | Google, ETH Zürich, Tübingen | paper | ||||||||||
| 2026-07-02 | Leanstral 1.5 | Mistral | model | 119B | 6.5B | 256K | |||||||
| 2026-07-02 | OpenAI Proposes Giving US Government a ~5% Equity Stake (~$42.6B) via an Alaska-Permanent-Fund-Style Vehicle CNBC | OpenAI | news | ||||||||||
| 2026-07-01 | V-SPLADE | Naver | model | 250M | |||||||||
| 2026-07-01 | Procedural Memory Distillation: Online Reflection for Self-Improving Language Models | Salesforce | paper | ||||||||||
| 2026-07-01 | NVIDIA Opens AI Infrastructure to Capital Partners; Anchor Investor in $10B+ Helix Digital Infrastructure NVIDIA | NVIDIA | news | ||||||||||
| 2026-06-30 | Claude Sonnet 5 | Anthropic | model | 1M | 38 (v4) | ||||||||
| 2026-06-30 | openPangu 2.0 | Huawei | model | 505B | 18B | 512K | |||||||
| 2026-06-30 | ★ LongCat-2.0 | Meituan | model | 1.6T | 48B | 1M | 19 (v4) | ||||||
| 2026-06-30 | SkillOpt | Microsoft | paper | ||||||||||
| 2026-06-30 | GeneBench-Pro | OpenAI | eval | 129 | |||||||||
| 2026-06-30 | ★ Sarashina3 | SB Intuitions | model | 30T | |||||||||
| 2026-06-30 | InnoSpark 3.0 | SII | model | 35B (max) | 262.14K | ||||||||
| 2026-06-30 | Fable 5 Redeployed After Export-Control Suspension Lifted; Anthropic Proposes Cross-Lab Jailbreak-Severity Framework Anthropic | Anthropic | news | ||||||||||
| 2026-06-30 | Meituan Releases LongCat-2.0 (1.6T MoE) — Trained Entirely on Domestic Chinese Chips, No NVIDIA Reuters | Meituan | news | ||||||||||
| 2026-06-30 | Japan Picks Noetra (SoftBank/NEC/Sony/Honda) + AIST for National Physical-AI Foundation Model; PFN Seconding ~15 Engineers Japan Times | PFN | news | ||||||||||
| 2026-06-29 | TabFM | model | 400M | ||||||||||
| 2026-06-29 | Brain2Qwerty | Meta | paper | ||||||||||
| 2026-06-29 | DOPD: Dual On-policy Distillation | NUS | paper | ||||||||||
| 2026-06-29 | SenseNova-Vision 2 | SenseTime | model paper | 7B | |||||||||
| 2026-06-29 | Smooth Scaling Laws Hide Stepwise Token Learning | Xiaohongshu | paper | ||||||||||
| 2026-06-29 | Korea's ~₩1,350T (~$880B) AI/Chips Investment Drive: Naver Among Companies Committing ₩550T in AI Data Centers CNBC | Naver | news | ||||||||||
| 2026-06-28 | OSWorld 2.0 | HKU | eval | 108 | |||||||||
| 2026-06-26 | Otter Weather | University of Cambridge | model | ||||||||||
| 2026-06-26 | DSpark / DeepSpec 2 | DeepSeek | paper library | ||||||||||
| 2026-06-26 | ★ GPT-5.6 (Sol / Terra / Luna) | OpenAI | model | 1M | 47 (v4) | ||||||||
| 2026-06-26 | Agents-A1 | PJLab | model | ||||||||||
| 2026-06-26 | GPT-5.6 Sol/Terra/Luna Launch Restricted to ~20 US-Government-Approved Partners Under June 2 Executive Order Axios | OpenAI | news | ||||||||||
| 2026-06-25 | IBM Debuts World's First Sub-1nm (0.7nm) Chip: 3D Nanostack, ~100B Transistors, Up to 70% Better Efficiency vs 2nm IBM | IBM | news | ||||||||||
| 2026-06-25 | KRAFTON Launches Dedicated AI Business Unit to Drive Commercialization Seoul Economic Daily | KRAFTON | news | ||||||||||
| 2026-06-25 | Meta Hires Virtue AI Founders Bo Li, Dawn Song, and Sanmi Koyejo Into Superintelligence Labs and FAIR Axios | Meta | news | ||||||||||
| 2026-06-24 | OlmoEarth | Ai2 | model | 128M | |||||||||
| 2026-06-24 | EdgeBench | ByteDance | eval | 134 | |||||||||
| 2026-06-24 | CS2-10k | Reka | dataset | ||||||||||
| 2026-06-24 | Sarashina2.2-TTS 3 | SB Intuitions | model paper dataset | 800M | |||||||||
| 2026-06-24 | Anthropic Accuses Alibaba of Largest-Known Distillation Attack: 28.8M Claude Exchanges via ~25,000 Fake Accounts CNBC | Alibaba | news | ||||||||||
| 2026-06-24 | Computer Use Becomes a Native Tool in Gemini 3.5 Flash, Spanning Browser, Mobile, and Desktop Google | news | |||||||||||
| 2026-06-24 | OpenAI and Broadcom Unveil Jalapeño, OpenAI's First Custom Inference Chip — Designed to Tape-Out in ~9 Months Using OpenAI's Own Models OpenAI | OpenAI | news | ||||||||||
| 2026-06-23 | Doubao-Seed-2.1 | ByteDance | model | ||||||||||
| 2026-06-23 | Mistral OCR 4 | Mistral | model | ||||||||||
| 2026-06-23 | Internal Data Repetition Destroys Language Models | Stanford | paper | ||||||||||
| 2026-06-23 | Can Scale Save Us From Plasticity Loss in LLMs? | Zyphra | paper | ||||||||||
| 2026-06-22 | AudioCALM: Continuous Autoregressive Language Modeling for Universal Audio Generation | Alibaba | paper | ||||||||||
| 2026-06-22 | Qwen-AgentWorld 3 | Alibaba | model paper dataset | 35B | 3B | 262.14K | |||||||
| 2026-06-22 | SingGuard | Ant Group | model | 8B (max) | |||||||||
| 2026-06-22 | Reinforcement Learning Towards Broadly and Persistently Beneficial Models | OpenAI | paper | ||||||||||
| 2026-06-22 | BOLD (British Open-ended Learning and Discovery Lab) | University of Oxford | announcement | ||||||||||
| 2026-06-22 | PLaMo 3.0 Prime | PFN | model | 256K | |||||||||
| 2026-06-22 | SOFAIR | University College London | announcement | ||||||||||
| 2026-06-22 | SpaceXAI Signs Colossus 2 Compute Deal with Reflection: $150M a Month from July 2026 Through 2029 (~$6.3B Ceiling) Axios | Reflection AI | news | ||||||||||
| 2026-06-22 | SpaceX Inks Compute Deal with Reflection AI for Colossus 2 (Up to ~$6.3B Through 2029) TechCrunch | SpaceXAI | news | ||||||||||
| 2026-06-22 | Zhipu AI Market Cap Tops HK$1 Trillion (~US$128B) as GLM-5.2 Sends Shares Soaring SCMP | Z.ai | news | ||||||||||
| 2026-06-20 | LabVLA | Zhejiang University | model | ||||||||||
| 2026-06-19 | Unlimited-OCR | Baidu | model | ||||||||||
| 2026-06-18 | Amazon in Talks to Sell Trainium3 Chips Directly to External Data Centers, Breaking AWS-Only Distribution Bloomberg (via Yahoo) | Amazon | news | ||||||||||
| 2026-06-18 | Bell, Cohere, Hypertec and BUZZ HPC Sign ~$220M Canadian Sovereign-AI Compute Deal (2,304 Grace Blackwell GPUs) Newswire (Bell/Cohere PR) | Cohere | news | ||||||||||
| 2026-06-18 | DeepMind Talent Exodus: Gemini Co-Lead Noam Shazeer Joins OpenAI; Nobel Laureate John Jumper Joins Anthropic Days Later Fortune | news | |||||||||||
| 2026-06-18 | Laguna M.1 Released as Open Weights (Apache 2.0); Base and Post-Trained Checkpoints on Hugging Face Hugging Face | Poolside | news | ||||||||||
| 2026-06-17 | Tmax 3 | Ai2 | model paper dataset | 27B (max) | |||||||||
| 2026-06-17 | LifeSciBench | OpenAI | eval | 750 | |||||||||
| 2026-06-17 | RNGBench | PJLab | eval | ||||||||||
| 2026-06-17 | Skill-Guided Continuation Distillation for GUI Agents | StepFun | paper | ||||||||||
| 2026-06-17 | Xiaohongshu plans for Hong Kong IPO by year-end, targets US$70b valuation The Standard (via WSJ) | Xiaohongshu | news | ||||||||||
| 2026-06-16 | Qwen-Robot 3 | Alibaba | paper | ||||||||||
| 2026-06-16 | WikiProfile / Recall Is the Bottleneck | paper | |||||||||||
| 2026-06-16 | From Reasoning Traces to Reusable Modules | MBZUAI | paper | ||||||||||
| 2026-06-16 | Deployment Simulation | OpenAI | paper | ||||||||||
| 2026-06-16 | DeepSeek Closes First External Round: ¥51B (~$7.4B) at ~¥400B (~$56B) — Largest AI Raise in China's History; Liang Retains Control Caixin | DeepSeek | news | ||||||||||
| 2026-06-16 | SpaceX Agrees to Acquire Cursor for $60B All-Stock, Days After IPO (Pending Regulatory Approval) TechCrunch | SpaceXAI | news | ||||||||||
| 2026-06-16 | Upstage Previews 'Solar 2', Claiming External Benchmarks Near Claude Sonnet 4.6; Third-Gen Model Promised by Year-End The Elec | Upstage | news | ||||||||||
| 2026-06-16 | Z.ai GLM-5.2 Tops the AA Intelligence Index as the Highest-Scoring Open Model (51); MIT Open Weights Released Crypto Briefing | Z.ai | news | ||||||||||
| 2026-06-15 | MolmoMotion | Ai2 | model | 4.85B | |||||||||
| 2026-06-15 | QK-Normed MLA | Ant Group | paper | ||||||||||
| 2026-06-15 | Anatomy of Post-Training | Anthropic | paper | ||||||||||
| 2026-06-15 | The Culture Funnel / CultureMarkers | Cohere | paper | ||||||||||
| 2026-06-15 | daVinci-kernel 2 | SII | model paper | 14B (max) | |||||||||
| 2026-06-15 | Norm-Agnostic Residual Networks (NAG) | Zyphra | paper | ||||||||||
| 2026-06-15 | HCLTech to Buy 10.5% Stake in Sarvam AI for ~$150M (₹14.27B), Valuing It at $1.5B — Series B First Close ($234M Raised), Unicorn Status Reuters | Sarvam | news | ||||||||||
| 2026-06-13 | ★ GLM-5.2 | Z.ai | model | 744B | 40B | 1M | 34 (v4) | ||||||
| 2026-06-12 | olmo-eval | Ai2 | library | ||||||||||
| 2026-06-12 | HarnessX | Huawei | paper | ||||||||||
| 2026-06-12 | LoHoSearch | Meituan | eval | 544 | |||||||||
| 2026-06-12 | Kimi Code CLI | Moonshot AI | library | ||||||||||
| 2026-06-12 | ★ Kimi K2.7-Code | Moonshot AI | model | 1T | 32B | 262.14K | 26 (v4) | ||||||
| 2026-06-12 | US Export-Control Order Forces Global Suspension of Claude Fable 5 and Mythos 5 Over Claimed Safeguard Bypass Anthropic | Anthropic | news | ||||||||||
| 2026-06-12 | BAAI Conference Unveils 'Wujie' World-Model Suite: Physis-v0.1, RoboBrain Orca, Brainμ 1.0 Neuroscience FM, OpenComplex2.5 (Announced, Artifacts Pending) BAAI | BAAI | news | ||||||||||
| 2026-06-12 | Mistral in Early Talks to Raise ~€3B at ~€20B Valuation, Nearly Double Its Series C (Unclosed) TechCrunch | Mistral | news | ||||||||||
| 2026-06-12 | Moonshot Launches Kimi Work: Local Desktop Agent with 300-Sub-Agent Swarm on Kimi K2.6 MarkTechPost | Moonshot AI | news | ||||||||||
| 2026-06-11 | MOSS-Audio | Fudan University | model | ||||||||||
| 2026-06-11 | MiniMax Sparse Attention (MSA) | MiniMax | paper | ||||||||||
| 2026-06-11 | NexAU (Agent Universe) | Nex-AGI | library | ||||||||||
| 2026-06-11 | MA-ProofBench | OpenBMB | eval | 200 | |||||||||
| 2026-06-11 | Hy-Embodied-0.5-VLA | Tencent | model | ||||||||||
| 2026-06-11 | Zonos2 (ZONOS2) | Zyphra | model | ||||||||||
| 2026-06-11 | SpaceX Prices Largest IPO Ever at $135/Share; Trades on Nasdaq as SPCX TechCrunch | SpaceXAI | news | ||||||||||
| 2026-06-10 | FORT-Searcher | Renmin (RUC) | paper | ||||||||||
| 2026-06-10 | MiMoCode | Xiaomi | library | ||||||||||
| 2026-06-09 | ★ Claude Fable 5 | Anthropic | model | 50 (v4) | |||||||||
| 2026-06-09 | ★ Claude Mythos 5 | Anthropic | model | ||||||||||
| 2026-06-09 | North Mini Code | Cohere | model | 30B | 3B | 262.14K | 3 | ||||||
| 2026-06-09 | ★ DiffusionGemma | model | 25.2B | 3.8B | 262.14K | 10 (v4) | |||||||
| 2026-06-09 | ★ Nex-N2 | Nex-AGI | model | 397B | 17B | 28 (v4) | |||||||
| 2026-06-09 | Cohere Releases North Mini Code — Its First Developer-Focused Model, a 30B-A3B Apache-2.0 Agentic Coder Cohere | Cohere | news | ||||||||||
| 2026-06-09 | StepFun Seeks US$12 Billion Valuation in Hong Kong IPO After ~$2.5B Pre-IPO Round DigiTimes | StepFun | news | ||||||||||
| 2026-06-08 | ★ Apple Foundation Models 3 (AFM 3) | Apple | model | 20B (max) | |||||||||
| 2026-06-08 | Baichuan-M4 | Baichuan | paper | ||||||||||
| 2026-06-08 | Active Inference as Context Acquisition for AI Agents | TU Munich | paper | ||||||||||
| 2026-06-08 | AMD and Imperial to Collaborate on AI-Enabled Scientific Discovery and Sovereign AI Imperial College London | Imperial | news | ||||||||||
| 2026-06-08 | China's Moonshot AI Seeks $30 Billion Value in New Funding Talks — Third Round in Six Months Bloomberg | Moonshot AI | news | ||||||||||
| 2026-06-08 | OpenAI Confidentially Files for IPO (Last Valued at $852B), One Week After Anthropic TechCrunch | OpenAI | news | ||||||||||
| 2026-06-08 | Xiaomi MiMo-V2.5-Pro-UltraSpeed Breaks 1000 Tokens/s on a 1T-Parameter Model (with TileRT) Xiaomi MiMo | Xiaomi | news | ||||||||||
| 2026-06-05 | DaX (大象) | Alibaba | paper | ||||||||||
| 2026-06-05 | ★ Chronos-2 | Amazon | model | 5.45K | 12.42M | 1 | 120M | 8.19K | |||||
| 2026-06-04 | BioMysteryBench | Anthropic | eval | 99 | |||||||||
| 2026-06-04 | OPI-Struc (STELLA) | BAAI | dataset | 1 | |||||||||
| 2026-06-04 | Agents' Last Exam | UC Berkeley | eval | ||||||||||
| 2026-06-04 | ★ Nemotron 3 Ultra | NVIDIA | model | 1.35K | 550B | 55B | 1M | 12 | |||||
| 2026-06-04 | Nemotron 3.5 Content Safety | NVIDIA | model | 5.91K | 4B | ||||||||
| 2026-06-04 | SciCore-Omics | OpenBMB | model | 8 | 238 | 8B | |||||||
| 2026-06-04 | dots.tts 2 | Xiaohongshu | model paper | 2B | |||||||||
| 2026-06-03 | CyberGym-E2E | UC Berkeley | eval | 920 | |||||||||
| 2026-06-03 | Gemma 4 12B | model | 816.16K | 11.95B | 262.14K | 14 (v4) | |||||||
| 2026-06-03 | ChartNet | IBM | dataset | ||||||||||
| 2026-06-03 | OfficeComprehensionBenchmark | Microsoft | eval | 3 | 1.04K | 2 | |||||||
| 2026-06-03 | RHELM | Microsoft | eval | 1.3K | 7 | ||||||||
| 2026-06-03 | NVIDIA Buys Enterprise Model-Maker Kumo AI for at Least $400 Million The Information | NVIDIA | news | ||||||||||
| 2026-06-02 | MAI-Code-1-Flash | Microsoft | model | 137B | 262.14K | ||||||||
| 2026-06-02 | ★ MAI-Thinking-1 | Microsoft | model | 1T | 35B | 30T | 262.14K | ||||||
| 2026-06-02 | Zamba2-VL (Vision-Language) | Zyphra | model | 7B | |||||||||
| 2026-06-02 | Build 2026: Seven MAI Models Launched — MAI-Thinking-1 (1T/35B Reasoning), MAI-Code-1-Flash, Multimodal Stack Refresh Microsoft AI | Microsoft | news | ||||||||||
| 2026-06-02 | 独家|原华为盘古“90后少帅”离职创业,新公司已获1亿美元估值新融资 (Former Huawei Pangu lead Wang Yunhe's new company reportedly raises at a $100M valuation) Sina Tech | TokenRhythm | news | ||||||||||
| 2026-06-02 | Zamba2-VL Released: Hybrid SSM Vision-Language Models (1.2B / 2.7B / 7B) Zyphra | Zyphra | news | ||||||||||
| 2026-06-01 | Qwen3.7-Plus | Alibaba | model | 1M | 25 (v4) | ||||||||
| 2026-06-01 | ★ MiniMax-M3 | MiniMax | model | 428B | 23B | 1.05M | 19 | ||||||
| 2026-06-01 | ★ Cosmos 3 3 | NVIDIA | model paper dataset | ||||||||||
| 2026-06-01 | MWA & FALCON (Speech Forced Alignment) 2 | Technion | model paper | ||||||||||
| 2026-06-01 | Anthropic Confidentially Files for IPO at ~$965B Valuation, First Among AI Labs Fortune | Anthropic | news | ||||||||||
| 2026-06-01 | Nemotron 3 Ultra Announced at Computex Taipei — 550B/55B MoE, AAII 48, Ships June 4 on HuggingFace Artificial Analysis | NVIDIA | news | ||||||||||
| 2026-05-31 | SCOPE | University of Edinburgh | paper | ||||||||||
| 2026-05-31 | HakushoBench | NII | eval | 3 | 2.05K | ||||||||
| 2026-05-31 | τ₀-World Model | SII | model | ||||||||||
| 2026-05-29 | SchGen | Microsoft | model | 15 | 20B | 13.31K | |||||||
| 2026-05-29 | Grok Build 0.1 (model) | SpaceXAI | model | 256K | 27 (v4) | ||||||||
| 2026-05-29 | Universal Audio Tokenizer | Tencent | model | 4 | |||||||||
| 2026-05-29 | Trinity Is Moving to OpenMDW-1.1 Arcee AI Blog | Arcee | news | ||||||||||
| 2026-05-28 | ★ Claude Opus 4.8 | Anthropic | model | 42 (v4) | |||||||||
| 2026-05-28 | Ultra-FineWeb-L3 | OpenBMB | dataset | ||||||||||
| 2026-05-28 | Draft-OPD: On-Policy Distillation for Speculative Draft Models | PJLab | paper | ||||||||||
| 2026-05-28 | ★ Step-3.7-Flash | StepFun | model | 50.19K | 198B | 11B | 262.14K | 19 (v4) | |||||
| 2026-05-28 | Autonomous Agentic Data Engineering | Tencent | paper | ||||||||||
| 2026-05-28 | ByteDance Developing Custom CPU Chips to Support AI Rollout; Pursuing Both Arm and RISC-V Tracks Reuters | ByteDance | news | ||||||||||
| 2026-05-28 | Microsoft to Unveil Homegrown Coding Model + Image / Reasoning / Speech / Transcription Suite at Build 2026 The Information | Microsoft | news | ||||||||||
| 2026-05-28 | Mistral Chases AI Superintelligence to Counter U.S. Dominance WSJ | Mistral | news | ||||||||||
| 2026-05-28 | SKT Launches A.Biz Cowork Internal AI Agent (Beta) and AXMS 1.5 Platform Upgrade Seoul Economic Daily | SK Telecom | news | ||||||||||
| 2026-05-27 | Pruning and Distilling Mixture-of-Experts into Dense Language Models | KRAFTON, KAIST | paper | ||||||||||
| 2026-05-27 | Harness-Bench | Peking University | eval | ||||||||||
| 2026-05-27 | Sci-Base | PJLab | dataset | ||||||||||
| 2026-05-26 | MOSS-TTS v1.5 | Fudan University | model | ||||||||||
| 2026-05-26 | Granite Guardian 4.1 | IBM | model | 1.17K | 8B | ||||||||
| 2026-05-26 | The MiniMax-M2 Series: Technical Report | MiniMax | paper | ||||||||||
| 2026-05-26 | LocateAnything-3B | NVIDIA | model | 131.79K | 3B | ||||||||
| 2026-05-26 | DeepMind CEO Demis Hassabis: Humanity Has 'a Few Years' to Prepare for AGI; 'Foothills of the Singularity' Axios | news | |||||||||||
| 2026-05-25 | mKernel: Fast Multi-GPU, Multi-Node Fused Kernels | UC Berkeley | library | ||||||||||
| 2026-05-25 | LLaVA-OneVision-2 | NTU | model | ||||||||||
| 2026-05-25 | MiniCPM5-1B | OpenBMB | model | 137.34K | 1.08B | 131.07K | 9 (v4) | ||||||
| 2026-05-24 | Granite Switch 4.1 | IBM | model | 76 | 1.1K | 30B | 131.07K | ||||||
| 2026-05-24 | Samudra 2 | NYU, Open Athena | model | ||||||||||
| 2026-05-23 | ScaleAcross Explorer | Meta | paper | ||||||||||
| 2026-05-23 | Nemotron-Labs Diffusion | NVIDIA | model | 12.46K | 14B | ||||||||
| 2026-05-23 | BitCPM-CANN | OpenBMB | model | 9.42K | 7.17K | 8B | |||||||
| 2026-05-22 | ★ Intern-S2-Preview | PJLab | model | 809 | 6.73K | 36B | 131.07K | ||||||
| 2026-05-22 | Liang Wenfeng Commits to AGI Mission Over Near-term Commercialization as ¥70B (~$10B) Round Advances Bloomberg | DeepSeek | news | ||||||||||
| 2026-05-22 | Reflection to Supply AI Models for the US Department of Energy's Genesis Mission and National Laboratories Axios | Reflection AI | news | ||||||||||
| 2026-05-21 | Hunyuan Model Matrix Refresh: TurboS, T1, T1-Vision, and Hunyuan Voice (Top-8 Chatbot Arena, +50% T1-Vision Speed) KrASIA | Tencent | news | ||||||||||
| 2026-05-21 | Tencent Open-Sources Hy-MT2 Translation Family (1.8B / 7B / 30B-A3B) + IFMTBench Tencent Hunyuan | Tencent | news | ||||||||||
| 2026-05-20 | SEA-LION v4.5 | AI Singapore | model | 27B | 262.14K | ||||||||
| 2026-05-20 | ★ Qwen3.7-Max-Preview | Alibaba | model | 1M | 29 (v4) | ||||||||
| 2026-05-20 | ★ Command A+ | Cohere | model | 113.99K | 218B | 25B | 128K | 6 | |||||
| 2026-05-20 | Lens | Microsoft | model | 234 | 3.93K | 3.8B | |||||||
| 2026-05-20 | Introducing the SEA-LION v4.5 Suite: Agentic Power and Speed (Qwen3.6-27B and Gemma 4 E2B Bases, Block-Diffusion Speculative Decoder) AI Singapore | AI Singapore | news | ||||||||||
| 2026-05-20 | Qwen3.7-Max-Preview Unveiled at Alibaba Cloud Summit — AA Intelligence Index 57 (#1 Among Chinese Labs) Qwen | Alibaba | news | ||||||||||
| 2026-05-20 | Cohere Releases Command A+ — First Full Apache-2.0 Open Model with Lossless 4-bit Quantization and Native Citations VentureBeat | Cohere | news | ||||||||||
| 2026-05-20 | Kakao Partners with Google DeepMind on SynthID for Kanana Models — First Asian Firm to Adopt Seoul Economic Daily | Kakao | news | ||||||||||
| 2026-05-20 | Q1 FY27 Earnings: $81.6B Revenue (+85% YoY), $80B Buyback Authorized, Dividend Raised 25x to $0.25 NVIDIA | NVIDIA | news | ||||||||||
| 2026-05-19 | Antigravity 2.0 | model | |||||||||||
| 2026-05-19 | ★ Gemini 3.5 Flash | model | 1M | 34 (v4) | |||||||||
| 2026-05-19 | Gemini Omni | model | |||||||||||
| 2026-05-19 | AI21 Lays Off 110 of Its 180 Employees and Halts Jamba Model Development; Nebius Deal Ends as a Commercial Partnership Around the Maestro Agent Platform Calcalist | AI21 Labs | news | ||||||||||
| 2026-05-19 | Anthropic Hires Andrej Karpathy to Pre-training Team; Will Lead Sub-team Using Claude to Accelerate Pretraining Research CNBC | Anthropic | news | ||||||||||
| 2026-05-19 | Hitachi Deploys Claude to ~290,000 Employees; Embeds in Lumada 3.0 / HMAX for Critical Infrastructure Hitachi | Anthropic | news | ||||||||||
| 2026-05-19 | I/O 2026: Gemini 3.5 Flash (AAII 55), Gemini Omni Video Generator, Antigravity 2.0, Gemini Spark, AI Ultra Repriced $200/mo Google | news | |||||||||||
| 2026-05-19 | Four years after ChatGPT, Xiaohongshu's AI restraint gives way to urgency KrASIA | Xiaohongshu | news | ||||||||||
| 2026-05-18 | ZEDA: Post-Trained MoE Can Skip Half Experts via Self-Distillation | PJLab | paper | ||||||||||
| 2026-05-18 | First Vera CPU Deliveries to Anthropic, OpenAI, SpaceXAI, and Oracle NVIDIA | NVIDIA | news | ||||||||||
| 2026-05-16 | Full Attention Strikes Back (RTPurbo) | Alibaba | paper | ||||||||||
| 2026-05-15 | Grok Build Launches — Agentic Coding CLI Competing with Claude Code, Codex, and Antigravity Engadget | SpaceXAI | news | ||||||||||
| 2026-05-14 | TWN: Think When Needed | Alibaba | paper | ||||||||||
| 2026-05-14 | JT-35B-Flash | China Mobile (Jiutian) | model | 35B | 256K | 19 (v4) | |||||||
| 2026-05-14 | VPD: Learning from Language Feedback via Variational Policy Distillation | Salesforce | paper | ||||||||||
| 2026-05-14 | Realtime Voice API GA + gpt-realtime-2 Family (3 New Audio Models) OpenAI | OpenAI | news | ||||||||||
| 2026-05-14 | HCLTech to Anchor $300M Sarvam Round at $1.5B; Bessemer +$50M; NVIDIA, Prosperity7 Participating Outlook Business | Sarvam | news | ||||||||||
| 2026-05-14 | SKT × Korean Defense Ministry Sign MOU on Applying Sovereign AI Foundation Model to Defense SK Telecom | SK Telecom | news | ||||||||||
| 2026-05-14 | SpaceXAI Division Bleeding Researchers Since Merger; 11+ to Meta, 7+ to Thinking Machines TechCrunch | SpaceXAI | news | ||||||||||
| 2026-05-13 | ExploitBench | CMU | eval | 41 | |||||||||
| 2026-05-13 | Granite Embedding Multilingual R2 | IBM | paper | ||||||||||
| 2026-05-13 | EMO: Frustratingly Easy Progressive Training of Extendable MoE | MBZUAI | paper | ||||||||||
| 2026-05-13 | NexRL | Nex-AGI | library | ||||||||||
| 2026-05-12 | Fara 1.5 | Microsoft | model | 27B (max) | |||||||||
| 2026-05-12 | OpenSquilla | TokenRhythm | library | ||||||||||
| 2026-05-12 | Kuaishou Plans to Spin Off Kling AI Video Unit at \$20B Valuation; Tencent in Talks for \$2B Pre-IPO Round The Information | Kuaishou | news | ||||||||||
| 2026-05-11 | ExploitGym | UC Berkeley, Anthropic, OpenAI, Google | eval | 898 | |||||||||
| 2026-05-11 | Delphi Open Scaling Suite | Open Athena, Stanford | model | ||||||||||
| 2026-05-11 | MiniCPM-V 4.6 | OpenBMB | model | 615.51K | 1.3B | 262.14K | 6 (v4) | ||||||
| 2026-05-11 | DeepSeek First External Funding Round Reportedly Near Close at $45–50B Valuation, Led by China's 'Big Fund III' SCMP | DeepSeek | news | ||||||||||
| 2026-05-11 | Zyphra Announces 15 MW of AMD Instinct MI355X GPU Capacity for Zyphra Cloud Memeburn | Zyphra | news | ||||||||||
| 2026-05-09 | SlimQwen: Pruning and Distillation in Large MoE Pre-training | Alibaba, MBZUAI | paper | ||||||||||
| 2026-05-09 | ★ ERNIE 5.1 | Baidu | model | ||||||||||
| 2026-05-09 | Step-Audio-R1.1 (Realtime) Tops Big Bench Audio at 96.4%, Surpassing Grok Voice Agent Artificial Analysis | StepFun | news | ||||||||||
| 2026-05-07 | Cola DLM | ByteDance | paper | ||||||||||
| 2026-05-07 | AI Co-Mathematician | paper | |||||||||||
| 2026-05-07 | ZAYA1-74B-Preview | Zyphra | model | 74B | 4B | 15T | 262.14K | ||||||
| 2026-05-07 | OMAI Compute Cluster Goes Live — $152M NSF + Blackwell-Ultra Infrastructure for Open Science AI Ai2 | Ai2 | news | ||||||||||
| 2026-05-07 | Kakao Announces Kanana 2.5 — 150B Agent-Focused LLM at Q1 Earnings Call Korea Herald | Kakao | news | ||||||||||
| 2026-05-07 | Kimi Chatbot Maker Moonshot AI Valued at $20 Billion in Meituan-Led Round Bloomberg | Meituan | news | ||||||||||
| 2026-05-07 | Kimi Chatbot Maker Moonshot AI Valued at $20 Billion in Meituan-Led Round Bloomberg | Moonshot AI | news | ||||||||||
| 2026-05-07 | ZAYA1-74B-Preview: Scaling Pretraining on AMD (74B/4B MoE) Zyphra | Zyphra | news | ||||||||||
| 2026-05-06 | ★ ZAYA1-8B | Zyphra | model | 8B | 700M | ||||||||
| 2026-05-06 | DeepSeek in Talks for First-Ever Outside Round at $45B; Tencent + Big Fund III in Lead Group TechCrunch | DeepSeek | news | ||||||||||
| 2026-05-05 | ProgramBench | Meta | eval | 200 | |||||||||
| 2026-05-05 | TRIBE v2 (Brain Activity Foundation Model) | Meta | paper | 1 | |||||||||
| 2026-05-05 | iOS 27 to Let Users Swap in Claude, Gemini, and Others as Default Apple Intelligence Model Bloomberg | Apple | news | ||||||||||
| 2026-05-05 | GPT-5.5 Instant Becomes Default ChatGPT Model; 52.5% Fewer Hallucinated Claims vs 5.3 Instant OpenAI | OpenAI | news | ||||||||||
| 2026-05-04 | Horizon Length in LLM Agent Training | Microsoft | paper | ||||||||||
| 2026-05-04 | Zyphra Launches Zyphra Cloud & Zyphra Inference — Serverless Inference for Open Models, AMD-First Zyphra | Zyphra | news | ||||||||||
| 2026-05-03 | National Growth Fund Invests ₩560B in Upstage as Its Second Direct Investment; Series C Post-Money Reported at ~₩2T Seoul Economic Daily | Upstage | news | ||||||||||
| 2026-05-03 | Korea's National Growth Fund and SIF Approve KRW 560B (~$400M) Direct Equity in Upstage — First Software Co. Recipient Seoul Economic Daily | Upstage | news | ||||||||||
| 2026-05-01 | Expressivity of Local Attention in Transformers | ETH Zürich | paper | ||||||||||
| 2026-05-01 | Huawei's AI Chip Gains Ground as DeepSeek and Others Shift Away from Nvidia Financial Times | Huawei | news | ||||||||||
| 2026-04-30 | OlmPool: Cracks in the Foundation | Ai2 | paper | 7 | |||||||||
| 2026-04-30 | SenseTime Is Running Its New Model on Chinese Chips WIRED | SenseTime | news | ||||||||||
| 2026-04-30 | xAI Launches Grok 4.3 with Improved Agentic Performance and Lower Pricing Artificial Analysis | SpaceXAI | news | ||||||||||
| 2026-04-29 | ★ Granite 4.1 | IBM | model | 152 | 619.44K | 30B | 15T | 512K | 7 (v4) | ||||
| 2026-04-29 | ★ Mistral Medium 3.5 | Mistral | model | 400.07K | 128B | 256K | 5 | ||||||
| 2026-04-29 | Granite 4.1 Released: 3B/8B/30B Dense Models, 512K Context, 8B Matches Prior 32B MoE IBM Research | IBM | news | ||||||||||
| 2026-04-28 | Agentic Harness Engineering | Fudan University | paper | ||||||||||
| 2026-04-28 | ★ Laguna M.1 | Poolside | model | 225.8B | 23.4B | 30T | 262.14K | ||||||
| 2026-04-28 | Laguna XS.2 | Poolside | model | 33.4B | 3B | 30T | 262.14K | ||||||
| 2026-04-28 | ★ MiMo-V2.5-Pro | Xiaomi | model | 72.22K | 1.02T | 42B | 1M | 26 (v4) | |||||
| 2026-04-28 | Poolside Launches Laguna XS.2 (Open, Apache 2.0) and Laguna M.1 Agentic Coding Models Poolside | Poolside | news | ||||||||||
| 2026-04-24 | ★ DeepSeek-V4 3 | DeepSeek | model paper | 6.84M | 1.6T | 49B | 33T | 1M | 30 (v4) | ||||
| 2026-04-24 | Cohere to Acquire Germany's Aleph Alpha, Creating a Transatlantic AI Champion Financial Times | Cohere | news | ||||||||||
| 2026-04-24 | DeepSeek-V4 Released: 1.6T/49B MoE, First Frontier Model Trained Entirely on Huawei Ascend 950PR, MIT License DeepSeek | DeepSeek | news | ||||||||||
| 2026-04-23 | Ling-2.6 2 | Ant Group | model paper | 1T | 17 (v4) | ||||||||
| 2026-04-23 | Sapiens2 | Meta | model | 5B | |||||||||
| 2026-04-23 | ★ GPT-5.5 | OpenAI | model | ~9.7T | 1M | 38 (v4) | |||||||
| 2026-04-23 | ★ Hy3 | Tencent | model | 373 | 82.33K | 295B | 21B | 256K | 25 (v4) | ||||
| 2026-04-23 | Reflection CEO Confirms Latest Funding Round Closed at $25 Billion Pre-Money Valuation CNBC | Reflection AI | news | ||||||||||
| 2026-04-22 | ★ Qwen3.6 Open-Weight Models 2 | Alibaba | model | 3.54K | 10.41M | 35B | 3B | 262.14K | 21 (v4) | ||||
| 2026-04-22 | ★ LLaDA 2.0-Uni | Ant Group | model | 7.22K | 16B | ||||||||
| 2026-04-22 | Tencent and Alibaba in Talks to Invest in DeepSeek at $20B+ Valuation — First External Funding Bloomberg | DeepSeek | news | ||||||||||
| 2026-04-22 | Cloud Next '26: TPU 8t/8i Announced, Deep Research Max Agents, Chrome Auto Browse, Thinking Machines Lab Multi-Billion Deal 9to5Google | news | |||||||||||
| 2026-04-21 | SpaceX Strikes Deal for Right to Acquire Cursor for $60B Bloomberg | SpaceXAI | news | ||||||||||
| 2026-04-20 | BAR: Branch-Adapt-Route | Ai2 | paper | 150 | |||||||||
| 2026-04-20 | Qwen3.6-Max-Preview | Alibaba | model | 256K | 28 (v4) | ||||||||
| 2026-04-20 | ★ Kimi K2.6 | Moonshot AI | model | 1T | 32B | 262.14K | 27 (v4) | ||||||
| 2026-04-20 | Qwen3.6-Max-Preview: Alibaba's Most Powerful Model, #1 on Six Coding Benchmarks (AA Intelligence Index 52) Qwen | Alibaba | news | ||||||||||
| 2026-04-20 | Amazon Invests $5B More (Total $13B); Anthropic Commits $100B+ AWS Spend Over 10 Years, Secures Up to 5GW Compute Anthropic | Anthropic | news | ||||||||||
| 2026-04-17 | Qwen3.5-Omni | Alibaba | model | 256K | 20 (v4) | ||||||||
| 2026-04-17 | ★ Grok 4.3 | SpaceXAI | model | 25 (v4) | |||||||||
| 2026-04-17 | OPD (Rethinking On-Policy Distillation) | Tsinghua | paper | ||||||||||
| 2026-04-16 | ★ Claude Opus 4.7 | Anthropic | model | ~4T | 41 (v4) | ||||||||
| 2026-04-16 | LeapAlign: Post-Training Flow Matching Models at Any Generation Step | ByteDance | paper | ||||||||||
| 2026-04-16 | Prefill-as-a-Service: Cross-Datacenter KVCache for Next-Generation Models | Moonshot AI | paper | ||||||||||
| 2026-04-16 | Claude Opus 4.7 Released Anthropic | Anthropic | news | ||||||||||
| 2026-04-16 | ByteDance Recruits DeepSeek R1 Lead Author Daya Guo for Seed Agent Team SCMP | ByteDance | news | ||||||||||
| 2026-04-16 | DeepSeek R1 Lead Author Daya Guo Joins ByteDance Seed Amid Intensifying AI Talent War SCMP | DeepSeek | news | ||||||||||
| 2026-04-16 | DeepSeek V4 Imminent — 1T-Parameter MoE to Run Solely on Huawei Ascend 950PR Chips Dataconomy | DeepSeek | news | ||||||||||
| 2026-04-16 | How France's Mistral Built a $14 Billion AI Empire by Not Being American Forbes | Mistral | news | ||||||||||
| 2026-04-16 | GPT-Rosalind Launched for Life Sciences Drug Discovery OpenAI | OpenAI | news | ||||||||||
| 2026-04-15 | Revenue Run Rate Hits $30B; VCs Offer Up to $800B Valuation Axios | Anthropic | news | ||||||||||
| 2026-04-15 | Upstage Becomes Korea's First Generative AI Unicorn with $126M Series C Seoul Economic Daily | Upstage | news | ||||||||||
| 2026-04-14 | Lightning OPD: Efficient Post-Training for Large Reasoning Models | NVIDIA | paper | ||||||||||
| 2026-04-14 | Gemini Robotics-ER 1.6 Launched; Boston Dynamics Partnership for Industrial AI Boston Dynamics | news | |||||||||||
| 2026-04-14 | NAACP Sues xAI Over Memphis Colossus Data Center Pollution CNBC | SpaceXAI | news | ||||||||||
| 2026-04-13 | OpenAI Touts Amazon Alliance, Says Microsoft Has 'Limited Our Ability' to Reach Enterprise CNBC | OpenAI | news | ||||||||||
| 2026-04-13 | StepFun Unwinding Offshore Structure to Pave Way for HK IPO at Up to $10B Reuters | StepFun | news | ||||||||||
| 2026-04-12 | SoftBank/NEC/Honda/Sony Form JV for Trillion-Parameter Physical AI Model; $6.3B Government Backing Nikkei Asia | SB Intuitions | news | ||||||||||
| 2026-04-10 | Nexus: Common Minima for Better Generalization | ByteDance | paper | ||||||||||
| 2026-04-10 | Alibaba Token Hub Created: 5 AI Units Consolidated Under CEO Eddie Wu; RMB 380B ($53B) 3-Year Commitment SCMP | Alibaba | news | ||||||||||
| 2026-04-10 | Cohere in Advanced Merger Talks with Germany's Aleph Alpha Reuters | Cohere | news | ||||||||||
| 2026-04-10 | SK Telecom Partners with Rebellions and Arm for Sovereign AI Inference Infrastructure Rebellions | SK Telecom | news | ||||||||||
| 2026-04-10 | xAI Spending Pushed SpaceX to Nearly $5B Loss; CFO Anthony Armstrong Departs The Information | SpaceXAI | news | ||||||||||
| 2026-04-09 | Metis: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Models | Alibaba | paper | ||||||||||
| 2026-04-09 | HiFloat4 Format for LLM Pre-training on Ascend NPUs | Huawei | paper | ||||||||||
| 2026-04-09 | ★ EXAONE 4.5 | LG | model | 33B | 256K | 13 (v4) | |||||||
| 2026-04-09 | Efficient RL Training for LLMs with Experience Replay | Meta | paper | ||||||||||
| 2026-04-09 | Nebius in Talks to Acquire AI21 After the NVIDIA Deal Falls Through Calcalist | AI21 Labs | news | ||||||||||
| 2026-04-09 | EXAONE 4.5 Released — LG's First Open-Weight Vision-Language Model Korea Herald | LG | news | ||||||||||
| 2026-04-09 | Naver Shuts Down Clova X Chatbot; Pivots to Vertical AI Integrated into Search, Shopping, Finance Seoul Economic Daily | Naver | news | ||||||||||
| 2026-04-08 | ★ Muse Spark | Meta | model | 260K | 31 (v4) | ||||||||
| 2026-04-08 | Muse Spark Unveiled — First Model from Superintelligence Labs (Proprietary) Bloomberg | Meta | news | ||||||||||
| 2026-04-08 | Zhipu Hikes Prices Again as China AI Monetization Wave Quickens Bloomberg | Z.ai | news | ||||||||||
| 2026-04-07 | ★ Harrier | Microsoft | model | 387.22K | 27B (max) | ||||||||
| 2026-04-07 | ★ GLM-5.1 | Z.ai | model | 3.39K | 123.47K | 744B | 40B | 26 (v4) | |||||
| 2026-04-07 | Claude Mythos Withheld from Public Release; Project Glasswing Cybersecurity Consortium Launched with Apple and Google Fortune | Anthropic | news | ||||||||||
| 2026-04-07 | Ascend 950PR AI Chip in Production; 750K Units Planned for 2026; Alibaba, ByteDance, Tencent Place Massive Orders TrendForce | Huawei | news | ||||||||||
| 2026-04-07 | GLM-5.1 Open-Source Release Scores #3 on Code Arena (1530 Elo); Stock Surges 19% BuildFastWithAI | Z.ai | news | ||||||||||
| 2026-04-06 | AI Agent Traps | paper | |||||||||||
| 2026-04-06 | MedGemma 1.5 | model | 1.51K | 416.35K | 4B | ||||||||
| 2026-04-06 | Multi-GW Compute Partnership Expansion with Google Cloud and Broadcom TechCrunch | Anthropic | news | ||||||||||
| 2026-04-06 | NVIDIA Acquires SchedMD (Slurm Workload Manager); Draws Regulatory Scrutiny Reuters | NVIDIA | news | ||||||||||
| 2026-04-03 | SkVM | Shanghai Jiao Tong University | library | ||||||||||
| 2026-04-03 | Microsoft Announces $10B Japan AI Infrastructure Investment (2026-2029) WSJ | Microsoft | news | ||||||||||
| 2026-04-03 | TII Launches Falcon Perception — 600M-Parameter Open Multimodal Model for Grounding and Segmentation TII | TII | news | ||||||||||
| 2026-04-03 | Xiaomi Reveals MiMo-V2-Pro (1T Parameters), Approaching GPT-5.2 / Opus 4.6 Performance VentureBeat | Xiaomi | news | ||||||||||
| 2026-04-02 | ★ Qwen 3.6-Plus | Alibaba | model | 1M | 27 (v4) | ||||||||
| 2026-04-02 | ★ Trinity Large Thinking | Arcee | model | 10.66K | 512K | 11 (v4) | |||||||
| 2026-04-02 | ★ Gemma 4 | model | 31B | 4B (max) | 256K | 4 | |||||||
| 2026-04-02 | Raon-OpenTTS 2 | KRAFTON | model dataset | ||||||||||
| 2026-04-02 | ★ Raon-Speech 3 | KRAFTON | model paper | ||||||||||
| 2026-04-02 | Raon-VisionEncoder | KRAFTON | model | 400M | |||||||||
| 2026-04-02 | ★ MAI Multimodal Stack (Transcribe / Voice / Image) | Microsoft | model | ||||||||||
| 2026-04-02 | SWE-HERO | NVIDIA | paper | ||||||||||
| 2026-04-02 | Alibaba Unveils Third Closed-Source AI Model in Focus on Profit Bloomberg | Alibaba | news | ||||||||||
| 2026-04-02 | Arcee's New Open-Source Trinity Large Thinking Is the Rare Powerful U.S.-Made Model VentureBeat | Arcee | news | ||||||||||
| 2026-04-02 | Gemma 4 Open Models Released Google Developers Blog | news | |||||||||||
| 2026-04-02 | KRAFTON Launches Raon, Its First Open-Source AI Model Family PTI | KRAFTON | news | ||||||||||
| 2026-04-02 | Poolside's $2B Series C Collapses; CoreWeave Exits 2GW Texas Data Center (Project Horizon) DataCenterDynamics | Poolside | news | ||||||||||
| 2026-04-02 | Sarvam AI Nearing $300-350M Raise at $1.5B Valuation Led by Bessemer with Nvidia and Amazon Bloomberg | Sarvam | news | ||||||||||
| 2026-04-01 | Simple Self-Distillation for Code Generation | Apple | paper | ||||||||||
| 2026-04-01 | Scaling Reasoning Tokens via RL and Parallel Thinking | ByteDance | paper | ||||||||||
| 2026-04-01 | OpenHarness | HKU | library | ||||||||||
| 2026-04-01 | Tempo-6B | KAUST | model | ||||||||||
| 2026-04-01 | Procedural Knowledge at Scale Improves Reasoning | Meta | paper | ||||||||||
| 2026-04-01 | Speech LLMs as Contextual Reasoning Transcribers | Microsoft | paper | ||||||||||
| 2026-04-01 | ★ GLM-5V-Turbo | Z.ai | model | 744B | 40B | 28.5T | 202.75K | 23 (v4) | |||||
| 2026-04-01 | Moonshot AI Raising $1B at $18B Valuation; Working with CICC and Goldman Sachs on HK IPO Bloomberg | Moonshot AI | news | ||||||||||
| 2026-03-31 | Think-Anywhere | Alibaba | paper | 55 | |||||||||
| 2026-03-31 | ASI-Evolve | SII | paper | 726 | |||||||||
| 2026-03-31 | OpenAI Closes $122B Round at $852B Valuation OpenAI | OpenAI | news | ||||||||||
| 2026-03-31 | Zhipu's Losses Climb 60% After Chinese AI Rivalry Worsens Bloomberg | Z.ai | news | ||||||||||
| 2026-03-31 | Zhipu's Losses Climb 60% After Chinese AI Rivalry Worsens Bloomberg | Z.ai | news | ||||||||||
| 2026-03-30 | Mistral AI Raises $830M in Debt to Set Up a Data Center Near Paris TechCrunch | Mistral | news | ||||||||||
| 2026-03-28 | ★ daVinci-LLM | SII | model | 154 | 3B | ||||||||
| 2026-03-28 | ★ Falcon Perception | TII | model | 715 | 12.87K | 600M | |||||||
| 2026-03-28 | DeepSeek Before V4: Culture, Organization, and Liang Wenfeng's Unique Goals (English summary) LatePost (晚点) | DeepSeek | news | ||||||||||
| 2026-03-26 | Cohere Transcribe | Cohere | model | 551.93K | 2B | ||||||||
| 2026-03-26 | Intern-S1-Pro | PJLab | model | 1T | 22B | ||||||||
| 2026-03-26 | China's Moonshot AI Seeks Listing in Hong Kong Under Heightened Scrutiny WSJ | Moonshot AI | news | ||||||||||
| 2026-03-25 | ★ LongCat-Next | Meituan | model | 438 | 2.18K | 74B | 3B | ||||||
| 2026-03-25 | AVO: Agentic Variation Operators for Autonomous Evolutionary Search | NVIDIA | paper | ||||||||||
| 2026-03-25 | daVinci-MagiHuman | SII | model | ||||||||||
| 2026-03-25 | Alibaba Launches AI Model Task Force; Top Researcher Resigns The Information | Alibaba | news | ||||||||||
| 2026-03-25 | MiniMax-M2.7, GLM-5 at 1/3 Cost Latent Space | MiniMax | news | ||||||||||
| 2026-03-24 | DeepSeek's Latest Job Postings Highlight Pivot to Agentic AI Bloomberg | DeepSeek | news | ||||||||||
| 2026-03-23 | SkillRouter | Alibaba | paper | ||||||||||
| 2026-03-23 | Felis | ByteDance | paper | ||||||||||
| 2026-03-20 | HuatuoGPT-3 | CUHK | model | ||||||||||
| 2026-03-20 | LongCat-Flash-Prover | Meituan | model | 85 | 64 | 560B | 27B | ||||||
| 2026-03-20 | EnterpriseOps-Gym | ServiceNow | eval | ||||||||||
| 2026-03-19 | ★ Nemotron Cascade 2 | NVIDIA | model | 3B | 1M | 12 (v4) | |||||||
| 2026-03-19 | dots.mocr | Xiaohongshu | model | 3B | |||||||||
| 2026-03-18 | Path-Constrained Mixture-of-Experts | Apple | paper | ||||||||||
| 2026-03-18 | Qianfan-OCR | Baidu | model | 174.67K | 4B | ||||||||
| 2026-03-18 | ★ MiniMax-M2.7 | MiniMax | model | 23 (v4) | |||||||||
| 2026-03-18 | MiMo-V2-Omni | Xiaomi | model | 25 (v4) | |||||||||
| 2026-03-18 | ★ MiMo-V2-Pro | Xiaomi | model | 1T | 42B | 1M | 29 (v4) | ||||||
| 2026-03-18 | MiMo-V2-TTS | Xiaomi | model | ||||||||||
| 2026-03-18 | Chinese AI Developer Zhipu to Create New Unit for Product Development The Information | Z.ai | news | ||||||||||
| 2026-03-17 | PRISM: Demystifying Retention and Interaction in Mid-Training | IBM | paper | ||||||||||
| 2026-03-17 | ★ Pre-training LLM without Learning Rate Decay Enhances Supervised Fine-Tuning | SB Intuitions | paper | ||||||||||
| 2026-03-16 | Mixture-of-Depths Attention (MoDA) | ByteDance | paper | 267 | |||||||||
| 2026-03-16 | ★ Mistral Small 4 | Mistral | model | 45.71K | 119B | 6.5B | 256K | 4 | |||||
| 2026-03-16 | Attention Residuals 2 | Moonshot AI | paper library | 3.3K | |||||||||
| 2026-03-16 | CUBE: A Standard for Unifying Agent Benchmarks | ServiceNow | paper | ||||||||||
| 2026-03-16 | Bridging the Semantic Gap: Announcing the SEA-LION Embedding Suite AI Singapore | AI Singapore | news | ||||||||||
| 2026-03-15 | Scientific Judge 2 | Baidu | paper dataset | 405 | |||||||||
| 2026-03-15 | Kernel Design Agents + KernelWiki | MIT | library | ||||||||||
| 2026-03-13 | OpenSWE / daVinci-Env | SII | dataset | 187 | |||||||||
| 2026-03-12 | RoboBrain-Dex | BAAI | model | 41 | |||||||||
| 2026-03-12 | IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL | CMU, MBZUAI | paper | ||||||||||
| 2026-03-12 | IndexShare (IndexCache): Cross-Layer Index Reuse for Sparse Attention | Z.ai | paper | ||||||||||
| 2026-03-11 | ★ Nemotron 3 Super | NVIDIA | model | 745.73K | 120B | 12B | 1M | 13 (v4) | |||||
| 2026-03-10 | ★ Exclusive Self Attention | Apple | paper | ||||||||||
| 2026-03-10 | Med-Asagi | UTokyo | model | ||||||||||
| 2026-03-10 | Ai2 CEO Ali Farhadi Steps Down; Microsoft Hires Key Researchers GeekWire | Ai2 | news | ||||||||||
| 2026-03-09 | PostTrainBench | Tübingen | eval | ||||||||||
| 2026-03-09 | Anthropic Sues Trump Admin Over Pentagon AI Blacklist CNBC | Anthropic | news | ||||||||||
| 2026-03-09 | OpenAI Acquires Promptfoo for AI Agent Security OpenAI | OpenAI | news | ||||||||||
| 2026-03-08 | CLI-Anything | HKU | library | ||||||||||
| 2026-03-08 | Scalable Training of MoE Models with Megatron Core | NVIDIA | paper | ||||||||||
| 2026-03-06 | ★ Sarvam-105B | Sarvam | model | 12.57K | 105B | 10.3B | 128K | 9 (v4) | |||||
| 2026-03-06 | ★ Sarvam-30B | Sarvam | model | 30B | 2.4B | 32K | 7 (v4) | ||||||
| 2026-03-05 | ★ GPT-5.4 | OpenAI | model | ~2.2T | 1M | 39 (v4) | |||||||
| 2026-03-05 | GPT-5.4 Released with 1M Token Context OpenAI | OpenAI | news | ||||||||||
| 2026-03-04 | RIVER | PJLab | dataset | 10 | |||||||||
| 2026-03-01 | ★ OLMo Hybrid | Ai2 | model | 48.9K | 7B | 6T | |||||||
| 2026-03-01 | ★ LLM-jp-4 | NII, University College London | model | 4.22K | 33B (max) | 3.8B | 11.7T | 65.54K | |||||
| 2026-02-28 | AnyTouch2 / ToucHD 2 | BAAI | dataset paper | ||||||||||
| 2026-02-27 | Solaris | NYU | model | ||||||||||
| 2026-02-25 | MaxClaw | MiniMax | library | ||||||||||
| 2026-02-25 | ZUNA (EEG Foundation Model) | Zyphra | model | 380M | |||||||||
| 2026-02-25 | Motif Founder Junghwan Lim Profiled as Korean Startups Take On the Conglomerates in the 'AI Squid Game'; the Science Ministry Selects Motif for Its Sovereign-Model Contest Bloomberg (via The Japan Times) | Motif Technologies | news | ||||||||||
| 2026-02-25 | Tencent-Backed AI Startup StepFun Is Said to Plan Hong Kong IPO Bloomberg | StepFun | news | ||||||||||
| 2026-02-19 | ★ Gemini 3.1 Pro | model | 1M | 30 (v4) | |||||||||
| 2026-02-19 | Gemini 3.1 Pro Released, Ties #1 on AA Intelligence Index Google DeepMind | news | |||||||||||
| 2026-02-18 | Large-scale Online Deanonymization with LLMs | ETH Zürich | paper | ||||||||||
| 2026-02-17 | OLMix | Ai2 | paper | ||||||||||
| 2026-02-17 | Tiny Aya | Cohere | model | 1.79K | 3.35B | 5 (v4) | |||||||
| 2026-02-17 | ★ Grok-4.20 | SpaceXAI | model | 2M | 26 (v4) | ||||||||
| 2026-02-17 | Mercury 2 Released: Diffusion LLM with AA Index 33 at 1000 tok/s Inception Labs | Inception Labs | news | ||||||||||
| 2026-02-16 | ★ Qwen3.5 5 | Alibaba | model | 3.54K | 397B | 17B | 1M | 21 (v4) | |||||
| 2026-02-16 | WebWorld | Alibaba | model | 39 | 1.26K | 32B | |||||||
| 2026-02-16 | ★ Ling 2.5 | Ant Group | model | 215 | 1T | 1M | |||||||
| 2026-02-16 | ZoomBench | Ant Group | dataset | 155 | |||||||||
| 2026-02-15 | Optimal Batch Size Scheduling via Functional Scaling Laws | Meituan | paper | ||||||||||
| 2026-02-14 | ★ Doubao-Seed-2.0 | ByteDance | model | ||||||||||
| 2026-02-14 | Doubao-Seed-2.0 Family Launched (Pro / Lite / Mini / Code) TechNode | ByteDance | news | ||||||||||
| 2026-02-13 | Cohere's $240M Year Sets Stage for IPO TechCrunch | Cohere | news | ||||||||||
| 2026-02-12 | ★ MiniMax-M2.5 | MiniMax | model | 585 | 589.2K | 229B | 23 (v4) | ||||||
| 2026-02-12 | GEBench | StepFun | dataset | 54 | |||||||||
| 2026-02-12 | FireRed-Image-Edit | Xiaohongshu | model | ||||||||||
| 2026-02-12 | Xiaomi-Robotics-0 2 | Xiaomi | model paper | 4.7B | |||||||||
| 2026-02-12 | Anthropic Raises $30B Series G at $380B Valuation Anthropic | Anthropic | news | ||||||||||
| 2026-02-11 | Ming-Flash-Omni-2.0 | Ant Group | model | 2.66K | |||||||||
| 2026-02-11 | MiniCPM-SALA | OpenBMB | model | 9.42K | 6.33K | 1M | |||||||
| 2026-02-11 | ★ Step-3.5-Flash 3 | StepFun | model paper dataset | 2.08K | 325.86K | 196B | 11B | 256K | 17 (v4) | ||||
| 2026-02-11 | ★ GLM-5 3 | Z.ai | model paper | 3.39K | 102.79K | 744B | 40B | 28.5T | 28 (v4) | ||||
| 2026-02-11 | Slime: Asynchronous RL for Agentic Tasks | Z.ai | library | 6.07K | |||||||||
| 2026-02-10 | Nanbeige4.1-3B | Nanbeige | model | 4B | 262.14K | ||||||||
| 2026-02-09 | Protenix | ByteDance | model | 1.94K | |||||||||
| 2026-02-09 | InternAgent-1.5 | PJLab | paper | ||||||||||
| 2026-02-08 | ★ Data Darwinism / Darwin Corpora | SII | dataset | 154 | |||||||||
| 2026-02-07 | Seedance 2.0 | ByteDance | model | ||||||||||
| 2026-02-07 | FireRed-OpenStoryline | Xiaohongshu | library | ||||||||||
| 2026-02-06 | Baichuan-M3 2 | Baichuan | paper model | 245 | 1.84K | 235B | |||||||
| 2026-02-05 | ★ Claude Opus 4.6 | Anthropic | model | ~5.3T | 1M | 32 (v4) | |||||||
| 2026-02-05 | KernelGYM & Dr.Kernel | HKUST | model | ||||||||||
| 2026-02-05 | ★ Kling 3.0 2 | Kuaishou | model paper | ||||||||||
| 2026-02-05 | Claude Opus 4.6 Released with 1M Context Anthropic | Anthropic | news | ||||||||||
| 2026-02-04 | RationaleRM | Alibaba | dataset | ||||||||||
| 2026-02-03 | ★ MiniCPM-o 4.5 | OpenBMB | model | 25.58K | 203.83K | ||||||||
| 2026-02-03 | CL-bench: A Benchmark for Context Learning | Tencent, Fudan University | eval | 1899 | |||||||||
| 2026-02-03 | EchoJEPA | Toronto & Vector Institute | model | ||||||||||
| 2026-02-02 | WAXAL: African Language Speech Corpus | dataset | |||||||||||
| 2026-02-02 | ★ Kimi K2.5 2 | Moonshot AI | model paper | 2.02K | 1.64M | 32B | 23 (v4) | ||||||
| 2026-02-02 | daVinci-Agency | SII | model | 9 | |||||||||
| 2026-02-02 | SpaceX Acquires xAI at $1.25T Combined Valuation Fortune | SpaceXAI | news | ||||||||||
| 2026-02-01 | nanobot | HKU | library | ||||||||||
| 2026-01-30 | Keel: Post-LayerNorm Is Back | ByteDance | paper | ||||||||||
| 2026-01-30 | GLM-OCR | Z.ai | model | 900M | |||||||||
| 2026-01-29 | SenseNova-MARS 3 | SenseTime | model paper dataset | 114 | 74 | 32B (max) | |||||||
| 2026-01-28 | Qwen3-ASR | Alibaba | model | 1.7B (max) | |||||||||
| 2026-01-28 | ★ Trinity Large | Arcee | model | 523 | 398B | 13B | 17T | 512K | |||||
| 2026-01-28 | Trinity Mini / Nano | Arcee | model | 14.85K | 26B | 3B | 131K | ||||||
| 2026-01-28 | SDPO (Self-Distillation) | ETH Zürich | paper | ||||||||||
| 2026-01-28 | ACE-Step-1.5 | StepFun | model | 10.97K | |||||||||
| 2026-01-27 | ★ K2 Think V2 | MBZUAI | model | 1.28K | 70B | 262.14K | 11 (v4) | ||||||
| 2026-01-27 | LongCat-Flash-Lite 2 | Meituan | model paper | 2.08K | 11 (v4) | ||||||||
| 2026-01-27 | Mistral AI Surges Revenue 20-Fold to Over $400 Million ARR MLQ | Mistral | news | ||||||||||
| 2026-01-27 | Tencent Bets Its AI Future on 28-Year-Old From OpenAI Caixin | Tencent | news | ||||||||||
| 2026-01-26 | DeepPlanning | Alibaba | dataset | ||||||||||
| 2026-01-26 | ★ daVinci-Dev | SII | model | 28 | |||||||||
| 2026-01-26 | ★ Solar Pro 3 | Upstage | model | 102B | 12B | 128K | 8 (v4) | ||||||
| 2026-01-23 | ★ LongCat-Flash-Thinking-2601 2 | Meituan | model paper | 254 | 7.66K | 560B | 27B | ||||||
| 2026-01-22 | ★ ERNIE 5.0 2 | Baidu | model paper | 2 | 2.4T | 14 (v4) | |||||||
| 2026-01-22 | EvoCUA | Meituan | library | 323 | 4.12K | ||||||||
| 2026-01-21 | CorpusQA: A 10 Million Token Benchmark for Corpus-Level Analysis and Reasoning | Alibaba | eval | ||||||||||
| 2026-01-21 | The Flexibility Trap (JustGRPO) | Tsinghua | paper | ||||||||||
| 2026-01-20 | ★ Yuan 3.0 Ultra | Inspur | model | 236 | 35 | 1T | 68.8B | ||||||
| 2026-01-20 | Scale-RAE | NYU | model | ||||||||||
| 2026-01-20 | Step-3-VL-10B 2 | StepFun | model paper | 406 | 484.26K | 8 (v4) | |||||||
| 2026-01-16 | Thinking Machines Suffers Wave of Defections as Co-Founders Zoph and Metz Return to OpenAI; Soumith Chintala Named CTO Fortune | Thinking Machines | news | ||||||||||
| 2026-01-15 | Francis Bach Awarded 2026 ERC Grant on the Reliability of AI Inria | INRIA | news | ||||||||||
| 2026-01-15 | Tao Qin Elected 2025 ACM Fellow ACM | ZGCA | news | ||||||||||
| 2026-01-12 | Engram: Conditional Memory via Scalable Lookup | DeepSeek | paper | ||||||||||
| 2026-01-12 | Alphabet Hits $4T Market Cap CNBC | news | |||||||||||
| 2026-01-11 | Distributional Clarity: The Hidden Driver of RL-Friendliness in Large Language Models | Baidu | paper | ||||||||||
| 2026-01-11 | ★ Solar Open 100B | Upstage | model | 3.87K | 102B | 12B | 19.7T | 131.07K | 10 (v4) | ||||
| 2026-01-09 | Agnes-SeaLLM-8B | Sapiens AI | model | 8B | |||||||||
| 2026-01-09 | PaCoRe: Learning to Scale Test-Time Compute | StepFun | paper | 334 | 110 | ||||||||
| 2026-01-09 | Zhipu and MiniMax IPO ChinaTalk | MiniMax | news | ||||||||||
| 2026-01-09 | Zhipu and MiniMax IPO ChinaTalk | Z.ai | news | ||||||||||
| 2026-01-08 | Afri-MCQA | MBZUAI | eval | ||||||||||
| 2026-01-06 | One Sample to Rule Them All: Extreme Data Efficiency in Multidiscipline Reasoning with Reinforcement Learning | Shanghai Jiao Tong University, Alibaba | paper | ||||||||||
| 2026-01-06 | xAI Raises $20B Series E at $230B Valuation CNBC | SpaceXAI | news | ||||||||||
| 2026-01-05 | Yuan 3.0 Flash | Inspur | model | 187 | 8 | 40B | 3.7B | ||||||
| 2026-01-05 | ★ K-EXAONE | LG | model | 236B | 23B | 14 (v4) | |||||||
| 2026-01-05 | ★ LFM2.5 | Liquid AI | model | 24B (max) | 2.45B (max) | 38T | 128K | 7 (v4) | |||||
| 2026-01-05 | ★ HyperCLOVA X SEED Omni | Naver | model | 648 | 8B | ||||||||
| 2026-01-05 | ★ Falcon-H1R | TII | model | 9 | 5.76K | 7B | 256K | 8 (v4) | |||||
| 2026-01-03 | ★ HyperCLOVA X SEED Think | Naver | model | 156.09K | 32B | 128K | 11 (v4) | ||||||
| 2026-01-01 | FlashInfer-python-paddle | Baidu | library | ||||||||||
| 2026-01-01 | Agentar-Z-100K | Z.ai | dataset | ||||||||||
| 2025-12-31 | FineWeb-Mask | ByteDance | dataset | ||||||||||
| 2025-12-31 | mHC: Manifold-Constrained Hyper-Connections | DeepSeek | paper | ||||||||||
| 2025-12-31 | OpenOneRec | Kuaishou | library | 812 | 33 | ||||||||
| 2025-12-30 | SeedFold | ByteDance | paper | ||||||||||
| 2025-12-30 | LongCat ZigZag Attention | Meituan | paper | 8 | 42 | ||||||||
| 2025-12-29 | KV-Tracker | Imperial | paper | ||||||||||
| 2025-12-27 | RollArt: Disaggregated Multi-Task Agentic RL Training at Scale | Alibaba | paper | ||||||||||
| 2025-12-27 | ★ A.X K1 | SK Telecom | model | 30 | 33.17K | 519B | 33B | 10T | 131K | ||||
| 2025-12-23 | ★ MiniMax-M2.1 | MiniMax | model | 544 | 10.19K | 229B | 21 (v4) | ||||||
| 2025-12-23 | VIBE & OctoCodingBench | MiniMax | dataset | ||||||||||
| 2025-12-23 | Step-DeepResearch | StepFun | library | 561 | |||||||||
| 2025-12-23 | Zhipu AI's Rise from Tsinghua Lab Pandaily | Z.ai | news | ||||||||||
| 2025-12-22 | SekoTalk / Seko 2.0 | SenseTime | model | 43 | |||||||||
| 2025-12-22 | ★ GLM-4.7 | Z.ai | model | 22 (v4) | |||||||||
| 2025-12-20 | TurboDiffusion | Tsinghua | library | ||||||||||
| 2025-12-19 | ★ Kanana-2 | Kakao | model | 150 | 30B | 3B | 32K | ||||||
| 2025-12-19 | Kakao Open-Sources Kanana-2 Model Optimized for Agentic AI Korea Times | Kakao | news | ||||||||||
| 2025-12-18 | ★ Seed1.8 | ByteDance | model | 218 | |||||||||
| 2025-12-18 | EXAONE Path 2.5 | LG | paper | ||||||||||
| 2025-12-18 | Towards Scalable Pre-training of Visual Tokenizers | MiniMax | paper | 490 | |||||||||
| 2025-12-18 | HY-Motion 1.0 | Tencent | paper | ||||||||||
| 2025-12-18 | Native Structured Latents for 3D Generation | Tsinghua | paper | ||||||||||
| 2025-12-18 | Seed1.8 Released as a Generalized Agentic Model ByteDance Seed | ByteDance | news | ||||||||||
| 2025-12-17 | Peter DeSantis to Lead Unified AGI Org; Rohit Prasad Departing CNBC | Amazon | news | ||||||||||
| 2025-12-17 | Tencent restructures AI operations, promotes high-profile recruit to chief AI scientist SCMP | Tencent | news | ||||||||||
| 2025-12-16 | ★ Molmo 2 | Ai2 | model | 641 | 1 | 8B (max) | |||||||
| 2025-12-16 | Motus | Tsinghua | model | ||||||||||
| 2025-12-16 | ★ MiMo-V2-Flash 2 | Xiaomi | model paper | 1.33K | 70.62K | 309B | 15B | 27T | 22 (v4) | ||||
| 2025-12-16 | MOPD: Multi-Teacher On-Policy Distillation | Xiaomi | paper | 1.33K | |||||||||
| 2025-12-15 | ★ Nemotron 3 Nano | NVIDIA | model | 2.18M | 30B | 3.5B | 25T | 1M | 9 (v4) | ||||
| 2025-12-15 | SonicMoE | Princeton | library | ||||||||||
| 2025-12-15 | NVIDIA in Advanced Talks to Acquire AI21 Labs for $2-3B SiliconANGLE | AI21 Labs | news | ||||||||||
| 2025-12-10 | ★ LLaDA 2 2 | Ant Group | model paper | 430 | 9.38K | ||||||||
| 2025-12-09 | ★ JAIS 2 | MBZUAI | model | 2.5K | 70B | 2.6T | 8.19K | ||||||
| 2025-12-08 | LongCat-Image 3 | Meituan | model paper | 695 | 47.59K | 6B | |||||||
| 2025-12-06 | ★ K2-V2 (LLM360) | MBZUAI | model | 183 | 70B | 10 (v4) | |||||||
| 2025-12-05 | NEO (Native VLM Architecture) 2 | SenseTime | model paper | 825 | 1 | 9B (max) | |||||||
| 2025-12-05 | ★ Hunyuan 2.0 | Tencent | model | 406B | 32B | 256K | |||||||
| 2025-12-04 | Nex-N1: Agentic Models via Large-Scale Environment Construction | Nex-AGI | paper | ||||||||||
| 2025-12-04 | Light-X | NTU | model | ||||||||||
| 2025-12-03 | Terminus-KIRA | KRAFTON | library | ||||||||||
| 2025-12-02 | ★ Amazon Nova 2 | Amazon | model | 1M | 14 (v4) | ||||||||
| 2025-12-02 | ★ Mistral Large 3 | Mistral | model | 1.98K | 675B | 41B | 256K | 9 (v4) | |||||
| 2025-12-02 | Nova 2 Model Family and Nova Act GA at re:Invent 2025 TechCrunch | Amazon | news | ||||||||||
| 2025-12-02 | Anthropic Acquires Bun, Claude Code Hits $1B ARR Anthropic | Anthropic | news | ||||||||||
| 2025-12-02 | Thomson Reuters and Imperial Announce Frontier AI Research Lab, Targeting a Competitive Frontier LLM Imperial College London | Imperial | news | ||||||||||
| 2025-12-01 | Ministral 3 | Mistral | model | 782.22K | 14B (max) | 5 (v4) | |||||||
| 2025-12-01 | John Giannandrea to Retire; Amar Subramanya Named VP of AI Apple | Apple | news | ||||||||||
| 2025-11-30 | RetroChimera | Microsoft | model | ||||||||||
| 2025-11-30 | gelab-zero (STEP-GUI) | StepFun | library | 2.19K | 2.21K | ||||||||
| 2025-11-28 | ★ LFM2 (Liquid Foundation Models 2) | Liquid AI | model | 24B | 2.3B | 6 (v4) | |||||||
| 2025-11-27 | DeepSeek-Math-V2 2 | DeepSeek | model dataset | 1.59K | 384 | ||||||||
| 2025-11-27 | MultiBanana | UTokyo | eval | ||||||||||
| 2025-11-25 | CABS + DataConcept-128M | Tübingen | dataset | ||||||||||
| 2025-11-24 | ★ Claude 4.5 Opus | Anthropic | model | ~3.4T | 200K | 29 (v4) | |||||||
| 2025-11-24 | HunyuanOCR | Tencent | model | 1.65K | 347.7K | 1B | |||||||
| 2025-11-23 | GAM: General Agentic Memory | BAAI | library | ||||||||||
| 2025-11-20 | ★ OLMo 3 | Ai2 | model | 10.31K | 32B | 5.9T | 65.54K | 7 (v4) | |||||
| 2025-11-20 | AICC: A 7.3T AI-Ready Corpus Built by a Model-Based HTML Parser | PJLab | dataset | ||||||||||
| 2025-11-20 | HunyuanVideo-1.5 | Tencent | model | 4.47K | 2.34K | ||||||||
| 2025-11-20 | MiMo-Embodied: X-Embodied Foundation Model | Xiaomi | paper | 1.12K | |||||||||
| 2025-11-19 | LPLB (Linear-Programming Load Balancer) | DeepSeek | library | 505 | |||||||||
| 2025-11-19 | Step-Audio-R1 | StepFun | model | 673 | 222 | 33B | |||||||
| 2025-11-19 | Yann LeCun Departs Meta to Found AMI Labs CNBC | Meta | news | ||||||||||
| 2025-11-17 | SenseNova-SI (Spatial Intelligence) 3 | SenseTime | model paper dataset | 271 | 8B (max) | ||||||||
| 2025-11-15 | ★ Doubao Seed Code | ByteDance | model | 256K | 17 (v4) | ||||||||
| 2025-11-15 | Doubao Seed Code (Reasoning Coder) Hits AA Intelligence Index 34 Artificial Analysis | ByteDance | news | ||||||||||
| 2025-11-14 | Miloco (Xiaomi Local Copilot) | Xiaomi | library | 2.61K | |||||||||
| 2025-11-13 | M100 Chip | Baidu | announcement | ||||||||||
| 2025-11-13 | Murati's Thinking Machines in Funding Talks at $50 Billion Value Bloomberg | Thinking Machines | news | ||||||||||
| 2025-11-12 | ★ AlphaProof | paper | |||||||||||
| 2025-11-12 | Interview: Ant Group's Open Model Ambitions Interconnects | Ant Group | news | ||||||||||
| 2025-11-12 | Chinese AI Prodigy Luo Fuli Joins Xiaomi as Industry Competition for Talent Heats Up South China Morning Post (via Yahoo Finance) | Xiaomi | news | ||||||||||
| 2025-11-10 | kosong | Moonshot AI | library | 520 | |||||||||
| 2025-11-10 | RLVE: Scaling Up RL with Adaptive Verifiable Environments | University of Washington, Ai2 | paper | ||||||||||
| 2025-11-09 | ReProbe | MBZUAI | paper | ||||||||||
| 2025-11-06 | InfinityStar | ByteDance | model | ||||||||||
| 2025-11-06 | Cambrian-S | NYU | model | ||||||||||
| 2025-11-06 | Step-Audio-EditX | StepFun | model | 929 | 43.36K | 3B | |||||||
| 2025-11-05 | SoftBank and SB Intuitions launch Sarashina API for enterprise access to Japanese LLM SoftBank | SB Intuitions | news | ||||||||||
| 2025-11-03 | ★ HPLT 3 | University of Edinburgh | dataset | ||||||||||
| 2025-11-03 | ★ LongCat-Flash-Omni 2 | Meituan | model paper | 492 | 62 | 560B | 27B | ||||||
| 2025-11-03 | How Zyphra Went All-In on AMD, With Head of Model Training Quentin Anthony (Latent Space Podcast) Latent Space | Zyphra | news | ||||||||||
| 2025-11-01 | popEVE | Harvard | model | ||||||||||
| 2025-11-01 | LightX2V | SenseTime | library | 2.36K | |||||||||
| 2025-11-01 | Inception Labs Raises $56M Seed from Menlo, Andrew Ng, Karpathy Inception Labs | Inception Labs | news | ||||||||||
| 2025-10-31 | GATE | LG | paper | ||||||||||
| 2025-10-30 | ★ Emu3.5 | BAAI | model | 1.52K | 907 | ||||||||
| 2025-10-30 | Kimi Linear 2 | Moonshot AI | model paper | 1.4K | 48B | 3B | 7 (v4) | ||||||
| 2025-10-30 | ★ Nicheformer | TU Munich | model | ||||||||||
| 2025-10-29 | ★ Ouro | ByteDance | model | 87.71K | 2.6B | 7.7T | |||||||
| 2025-10-29 | Toolathlon (The Tool Decathlon) | HKUST | eval | 108 | |||||||||
| 2025-10-29 | ★ Gaperon | INRIA | model | 24B | 4T | ||||||||
| 2025-10-28 | ODesign | BAAI | model | 311 | |||||||||
| 2025-10-28 | URSA (Uniform Discrete Diffusion) | BAAI | model | ||||||||||
| 2025-10-28 | Parallel Loop Transformer | ByteDance | paper | ||||||||||
| 2025-10-28 | OpenAI Completes For-Profit PBC Restructuring OpenAI | OpenAI | news | ||||||||||
| 2025-10-27 | ★ MiniMax-M2 | MiniMax | model | 2.6K | 127.97K | 230B | 10B | 19 (v4) | |||||
| 2025-10-27 | CoKE: Context as the Key to Biomolecular Understanding | PJLab | paper | 18 | |||||||||
| 2025-10-27 | JanusCoder | PJLab | model | 80 | 47 | ||||||||
| 2025-10-27 | Hunyuan Mirror | Tencent | paper | 1.14K | 2.99K | 1 | |||||||
| 2025-10-27 | Artificial Hivemind | University of Washington | paper | ||||||||||
| 2025-10-26 | BoltzGen | MIT, Open Athena | model | ||||||||||
| 2025-10-25 | LongCat-Video 3 | Meituan | model paper | 4.26K | 3.21K | 13.6B | |||||||
| 2025-10-24 | Huxley-Gödel Machine | KAUST | paper | ||||||||||
| 2025-10-24 | KAT-Coder 3 | Kuaishou | model paper | 72B | 22 (v4) | ||||||||
| 2025-10-23 | Anthropic to Expand Google Cloud TPU Use to 1M+ TPUs Anthropic | Anthropic | news | ||||||||||
| 2025-10-23 | Aeronautics Spinout Monolith AI Acquired by CoreWeave Imperial College London | Imperial | news | ||||||||||
| 2025-10-22 | Seed3D 1.0 | ByteDance | model | ||||||||||
| 2025-10-22 | ★ Nanbeige4-3B | Nanbeige | model | 4B | 23T | 65.54K | |||||||
| 2025-10-20 | DeepSeek-OCR / OCR-2 | DeepSeek | model | 23.27K | 2.35M | ||||||||
| 2025-10-17 | LongCat-Audio-Codec | Meituan | paper | 301 | |||||||||
| 2025-10-16 | MorphoBench | ZGCA | paper | 13 | |||||||||
| 2025-10-15 | Claude Haiku 4.5 | Anthropic | model | 200K | 17 (v4) | ||||||||
| 2025-10-15 | ★ Granite 4.0 | IBM | model | 32B | 9B | 6 (v4) | |||||||
| 2025-10-15 | ★ ScaleRL: The Art of Scaling Reinforcement Learning Compute for LLMs | Meta, UT Austin | paper | ||||||||||
| 2025-10-15 | InteractiveOmni 2 | SenseTime | model paper | 8 | 8B (max) | ||||||||
| 2025-10-15 | Anthropic Releases Claude Haiku 4.5: Sonnet 4-Level Coding at One Third the Cost, $1/$5 per Million Tokens Anthropic | Anthropic | news | ||||||||||
| 2025-10-15 | Granite 4.0: Hybrid Mamba Architecture, First ISO 42001 Certified Open Models IBM | IBM | news | ||||||||||
| 2025-10-14 | Rex-Omni 2 | IDEA Lab | model paper | 1.44K | 33.78K | 3B | |||||||
| 2025-10-14 | HAL — Holistic Agent Leaderboard | Princeton | eval | ||||||||||
| 2025-10-14 | Zhipu AI Breaks US Chip Reliance With First Major Model Trained on Huawei Stack SCMP | Z.ai | news | ||||||||||
| 2025-10-13 | RITE: Reinforcement Learning for Tool-Integrated Interleaved Thinking | Meituan | paper | ||||||||||
| 2025-10-13 | RAE — Representation Autoencoders | NYU | paper | ||||||||||
| 2025-10-13 | Rollout Routing Replay (R3) | Xiaomi | paper | ||||||||||
| 2025-10-10 | StreamingVLM | MIT | model | ||||||||||
| 2025-10-10 | DSPO: Stable and Efficient Policy Optimization for Agentic Search and Reasoning | Sapiens AI | paper | ||||||||||
| 2025-10-09 | ★ Ling 2.0 / Ling-1T 2 | Ant Group | model paper | 3.35K | 1T | 50B | 9 (v4) | ||||||
| 2025-10-09 | Reflection Raises $2B Led by Nvidia at an $8B Valuation to Be America's Open Frontier AI Lab TechCrunch | Reflection AI | news | ||||||||||
| 2025-10-06 | Paper2Video | NUS | library | ||||||||||
| 2025-10-01 | R-HORIZON-Websearch | Meituan | dataset | 26 | |||||||||
| 2025-10-01 | Code2Video | NUS | library | ||||||||||
| 2025-10-01 | BroRL: Scaling Reinforcement Learning via Broadened Exploration | NVIDIA | paper | ||||||||||
| 2025-10-01 | GDPval | OpenAI | eval | 1 | 1320 | ||||||||
| 2025-10-01 | ★ Apriel 15B Thinker | ServiceNow | model | 131.07K | 13 (v4) | ||||||||
| 2025-10-01 | Tinker | Thinking Machines | library | ||||||||||
| 2025-10-01 | IBM Research Names Jay Gambetta as Director; Dario Gil to DOE IBM | IBM | news | ||||||||||
| 2025-09-30 | Knapsack RL: Unlocking Exploration of LLMs via Optimizing Budget Allocation | ByteDance | paper | ||||||||||
| 2025-09-30 | CritPt | UIUC | eval | 190 | 71 | ||||||||
| 2025-09-30 | ★ GLM-4.6 | Z.ai | model | 355B | 19 (v4) | ||||||||
| 2025-09-29 | Ring 4 | Ant Group | model paper | 258 | 38.2K | 1T | 63B | 262.14K | 17 (v4) | ||||
| 2025-09-29 | ★ DeepSeek-V3.2 2 | DeepSeek | model paper | 3.59M | 2 | 685B | 37B | 21 (v4) | |||||
| 2025-09-29 | MGM-Omni | HKUST | model | ||||||||||
| 2025-09-29 | Scaling Behaviors of LLM Reinforcement Learning Post-Training | PJLab | paper | ||||||||||
| 2025-09-29 | AIRoA MoMa | UTokyo | dataset | ||||||||||
| 2025-09-28 | LLaVA-OneVision-1.5 | NTU | model | ||||||||||
| 2025-09-28 | HunyuanImage-3.0 2 | Tencent | model paper | 3.12K | 2.61K | 13B | |||||||
| 2025-09-26 | Qwen3Guard | Alibaba | model | 465 | |||||||||
| 2025-09-25 | Expanding Reasoning Potential (CoTP) | Meituan | paper | ||||||||||
| 2025-09-24 | LRM-Eval / ROME | BAAI | dataset | 5 | |||||||||
| 2025-09-23 | ByteWrist | ByteDance | model | ||||||||||
| 2025-09-23 | ★ LongCat-Flash-Thinking 2 | Meituan | model paper | 285 | 103 | 560B | 27B | ||||||
| 2025-09-23 | Symphony-MoE | PCL | paper | ||||||||||
| 2025-09-22 | BGE-Reasoner | BAAI | model | 31 | 959 | ||||||||
| 2025-09-22 | ScaleCUA | PJLab | model | 1.11K | 58 | ||||||||
| 2025-09-18 | Seedream 4.0 | ByteDance | model | 1 | |||||||||
| 2025-09-17 | ★ AToken | Apple | paper | 140 | |||||||||
| 2025-09-16 | Shanghai launches innovation institute to bridge AI research and industry Shanghai Municipal Government | SII | news | ||||||||||
| 2025-09-15 | checkpoint-engine | Moonshot AI | library | 963 | |||||||||
| 2025-09-10 | AgentGym-RL | Fudan University | paper | ||||||||||
| 2025-09-10 | Connectionism (research blog) | Thinking Machines | blog | ||||||||||
| 2025-09-08 | ★ mmBERT | Johns Hopkins | model | 307M | 3T | 8.19K | |||||||
| 2025-09-08 | ★ PLaMo 2 | PFN | model | 34.05K | 31B | 2T | 32K | ||||||
| 2025-09-05 | Klear 3 | Kuaishou | model paper | 82 | 58 | 1 | 46B | 2.5B | |||||
| 2025-09-05 | ★ MiniCPM4.1 2 | OpenBMB | model paper | 9.42K | 49.98K | 8B | |||||||
| 2025-09-02 | Baichuan-M2 2 | Baichuan | paper model | 212 | 938 | 1 | 32B | ||||||
| 2025-09-02 | Fantastic Pretraining Optimizers | Stanford | paper | ||||||||||
| 2025-09-02 | ★ Apertus | Swiss AI | model | 161.34K | 2 | 70B | 15T | 65.54K | 5 (v4) | ||||
| 2025-09-02 | DynaGuard | University of Maryland | model | ||||||||||
| 2025-09-02 | Hugo Larochelle named Mila Scientific Director Mila | Mila | news | ||||||||||
| 2025-09-01 | VeOmni | ByteDance | library | 2K | |||||||||
| 2025-09-01 | Benchmarking Optimizers for LLM Pretraining | EPFL | paper | ||||||||||
| 2025-09-01 | Dream-Coder 7B & DreamOn | HKU | model | ||||||||||
| 2025-09-01 | ★ LongCat-Flash-Chat 2 | Meituan | model paper | 1.34K | 81.37K | 1 | 560B | 27B | 128K | ||||
| 2025-09-01 | PlantCAD2 and GeneCAD | Open Athena | model | 694M | 8.19K | ||||||||
| 2025-09-01 | Hunyuan-MT | Tencent | model | 710 | 56.53K | 30B | 3B | ||||||
| 2025-09-01 | RLinf | ZGCA | library | ||||||||||
| 2025-09-01 | TwinBrainVLA | ZGCA | paper | ||||||||||
| 2025-09-01 | Darwin Monkey | Zhejiang University | announcement | ||||||||||
| 2025-09-01 | Mila unveils CAD $250M LaSalle Sovereign AI Research Hub Mila | Mila | news | ||||||||||
| 2025-09-01 | Mistral AI Raises EUR 2B at EUR 12B Valuation Mistral AI | Mistral | news | ||||||||||
| 2025-08-28 | HyperOS 3 | Xiaomi | announcement | ||||||||||
| 2025-08-26 | ★ MiniCPM-V 4.5 2 | OpenBMB | model paper | 25.58K | 93.36K | ||||||||
| 2025-08-25 | SEA-LION v4 | AI Singapore | model | 32B (max) | 500B | 131.07K | |||||||
| 2025-08-25 | GEPO | PCL | paper | ||||||||||
| 2025-08-25 | InternVL 3.5 | PJLab | model | 10.06K | 3 | 241B (max) | 28B (max) | ||||||
| 2025-08-25 | SEA-LION v4: Our First Multimodal Release (Gemma-SEA-LION-v4-27B-IT, #5 of 55 on SEA-HELM) AI Singapore | AI Singapore | news | ||||||||||
| 2025-08-23 | HunyuanVideo-Foley | Tencent | paper | ||||||||||
| 2025-08-21 | Fin-PRM: Process Reward Model for Financial Reasoning | Alibaba | paper | 502 | |||||||||
| 2025-08-21 | Waver | ByteDance | model | 938 | |||||||||
| 2025-08-21 | ★ DeepSeek-V3.1 | DeepSeek | model | 14 (v4) | |||||||||
| 2025-08-21 | Intern-S1 | PJLab | model | 241B | 28B | ||||||||
| 2025-08-20 | Seed-OSS-36B | ByteDance | model | 885 | 36.95K | 36B | 12T | 512K | 12 (v4) | ||||
| 2025-08-20 | ShizhenGPT | CUHK | model | ||||||||||
| 2025-08-20 | Nemotron Nano V2 | NVIDIA | model | 15.37K | 12B | 128K | |||||||
| 2025-08-20 | Seed-OSS-36B Released as Apache-2.0 Open-Weight Model VentureBeat | ByteDance | news | ||||||||||
| 2025-08-19 | NovoMolGen | Mila | model | ||||||||||
| 2025-08-15 | PXDesign | ByteDance | model | 229 | |||||||||
| 2025-08-15 | Physical Autoregressive Model (PAR) | PCL | paper | ||||||||||
| 2025-08-14 | NextStep-1 2 | StepFun | model paper | 688 | 62 | 14B | |||||||
| 2025-08-14 | Hunyuan-GameCraft 1.0 | Tencent | model | 723 | 76 | ||||||||
| 2025-08-14 | Cohere Raises $500M at $6.8B Valuation Cohere | Cohere | news | ||||||||||
| 2025-08-14 | Cohere Hires Long-Time Meta Research Head Joelle Pineau as Chief AI Officer TechCrunch | Cohere | news | ||||||||||
| 2025-08-12 | OpenCUA | HKU | model | ||||||||||
| 2025-08-12 | Mistral Medium 3.1 | Mistral | model | 128K | 9 (v4) | ||||||||
| 2025-08-12 | InternBootcamp | PJLab | library | 349 | |||||||||
| 2025-08-11 | GLM-4.5V | Z.ai | model | 2.33K | 167.01K | 106B | 12B | 64K | 8 (v4) | ||||
| 2025-08-07 | LLMEval-Fair | Fudan University | eval | 200K | |||||||||
| 2025-08-07 | CANN | Huawei | library | ||||||||||
| 2025-08-07 | ★ GPT-5 | OpenAI | model | ~4.1T | 400K | 23 (v4) | |||||||
| 2025-08-07 | TMA-Adaptive FP8 Grouped GEMM | PJLab | paper | 25 | |||||||||
| 2025-08-06 | ACAVCaps | Xiaomi | dataset | 424 | |||||||||
| 2025-08-05 | OmniScale | ByteDance | paper | 2K | |||||||||
| 2025-08-05 | Seed Diffusion | ByteDance | model | ||||||||||
| 2025-08-05 | ★ gpt-oss | OpenAI | model | 117B | 5.1B | 131.07K | 12 (v4) | ||||||
| 2025-08-05 | dots.vlm1 | Xiaohongshu | model | ||||||||||
| 2025-08-05 | OpenAI Releases gpt-oss-120b and gpt-oss-20b, Its First Open-Weight Language Models Since GPT-2 (Apache 2.0) Hugging Face | OpenAI | news | ||||||||||
| 2025-08-01 | Qwen-Image 5 | Alibaba | model | 7.98K | 173.33K | 20B | |||||||
| 2025-08-01 | MegaDFT | ZGCA | paper | ||||||||||
| 2025-08-01 | Ai2 and UW Awarded $152M from NSF and NVIDIA for Open Scientific AI GeekWire | Ai2 | news | ||||||||||
| 2025-07-31 | Seed-Prover | ByteDance | model | 433 | |||||||||
| 2025-07-30 | dots.ocr | Xiaohongshu | model | 3B | |||||||||
| 2025-07-29 | Libra-Bench & PIE_bench | Meituan | dataset | ||||||||||
| 2025-07-28 | SmallThinker | Shanghai Jiao Tong University | model | 21B | 3B | ||||||||
| 2025-07-28 | MixGRPO | Tencent | paper | 1.14K | |||||||||
| 2025-07-28 | ★ GLM-4.5 2 | Z.ai | model paper | 2 | 355B | 13 (v4) | |||||||
| 2025-07-27 | SenseNova V6.5 | SenseTime | model | ||||||||||
| 2025-07-27 | StepFun-Prover-Preview | StepFun | model | 35 | 60 | 32B | |||||||
| 2025-07-27 | HunyuanWorld 3 | Tencent | model | 2.85K | 615 | ||||||||
| 2025-07-26 | ARPO (Agentic Reinforced Policy Optimization) | Renmin (RUC) | paper | ||||||||||
| 2025-07-25 | ★ Step-3 2 | StepFun | model paper | 453 | 144.09K | 321B | 38B | ||||||
| 2025-07-24 | A.X 3.1 | SK Telecom | model | 13 | 305 | 34B | |||||||
| 2025-07-24 | SoftBank Corp. to Build the World's Largest AI Computing Infrastructure Using NVIDIA DGX SuperPOD with NVIDIA Blackwell GPUs SB Intuitions | SB Intuitions | news | ||||||||||
| 2025-07-23 | Towards Greater Leverage: Scaling Laws for Efficient MoE | Ant Group | paper | ||||||||||
| 2025-07-23 | ASI-Arch | SII | paper | 726 | |||||||||
| 2025-07-22 | Qwen-Code | Alibaba | library | 25.07K | |||||||||
| 2025-07-22 | Qwen3-Coder 2 | Alibaba | model | 16.61K | 1M | 480B (max) | 35B (max) | 10 (v4) | |||||
| 2025-07-22 | Seed-X Series | ByteDance | model | 172 | 130 | 7B | |||||||
| 2025-07-22 | Reka Raises $110M Series B at $1B Valuation Reka | Reka | news | ||||||||||
| 2025-07-17 | Agentar-DeepFinance-100K | Ant Group | dataset | 35 | |||||||||
| 2025-07-17 | Apple Intelligence Foundation Models Tech Report 2025 Apple | Apple | news | ||||||||||
| 2025-07-15 | Ettin | Johns Hopkins | model | 1B | 2T | ||||||||
| 2025-07-15 | Mira Murati's Thinking Machines Lab Is Worth $12B in Seed Round TechCrunch | Thinking Machines | news | ||||||||||
| 2025-07-14 | HKGAI-V1 | HKUST | model | ||||||||||
| 2025-07-14 | Mixture-of-Recursions | KAIST | paper | ||||||||||
| 2025-07-14 | ★ EXAONE 4.0 | LG | model | 33.89K | 32B | 8 (v4) | |||||||
| 2025-07-12 | Scaling Laws for Optimal Data Mixtures | Apple | paper | ||||||||||
| 2025-07-11 | ★ Kimi K2 4 | Moonshot AI | model paper | 10.84K | 2.71M | 5 | 1T | 32B | 256K | 13 (v4) | |||
| 2025-07-10 | FlexOlmo | Ai2 | model | 150 | 1.4K | 33B | |||||||
| 2025-07-10 | H-Net | CMU | model | ||||||||||
| 2025-07-10 | KAT (Kwai-AutoThink) 2 | Kuaishou | paper model | 55 | 72B | ||||||||
| 2025-07-09 | EXAONE Path 2.0 | LG | paper | ||||||||||
| 2025-07-09 | CADmium | Mila | model | ||||||||||
| 2025-07-09 | ★ Grok-4 | SpaceXAI | model | ~3.2T | 256K | 22 (v4) | |||||||
| 2025-07-09 | Grok-4 Released with Native Tool Use and Reasoning xAI | SpaceXAI | news | ||||||||||
| 2025-07-08 | ★ NeoBabel | University of Amsterdam | model | ||||||||||
| 2025-07-07 | POLAR | PJLab | paper | 167 | |||||||||
| 2025-07-05 | How to Train Your LLM Web Agent | ServiceNow | paper | ||||||||||
| 2025-07-03 | ★ IFBench | Ai2 | eval | 140 | 1 | 300 | 58 | ||||||
| 2025-07-03 | ★ A.X 4.0 | SK Telecom | model | 158 | 653 | ||||||||
| 2025-07-01 | CodePRM: Execution Feedback-enhanced Process Reward Model for Code Generation | Huawei | paper | ||||||||||
| 2025-07-01 | Voxtral | Mistral | model | 337.47K | 4B (max) | ||||||||
| 2025-07-01 | mini-swe-agent | Princeton | library | ||||||||||
| 2025-07-01 | ★ Solar Pro 2 | Upstage | model | 31B | 64K | 7 (v4) | |||||||
| 2025-06-30 | ★ openPangu | Huawei | announcement | ||||||||||
| 2025-06-30 | Meta Superintelligence Labs Created; Wang Named Chief AI Officer CNBC | Meta | news | ||||||||||
| 2025-06-27 | ★ HyperCLOVA X THINK | Naver | model | 128K | |||||||||
| 2025-06-27 | Hunyuan-A13B 2 | Tencent | model paper | 816 | 46.54K | 80B | 13B | 256K | |||||
| 2025-06-26 | Kwai Keye-VL 3 | Kuaishou | model paper | 785 | 192.8K | 31B | 3B | 262.14K | |||||
| 2025-06-25 | OctoThinker | SII | model | 188 | 8B (max) | ||||||||
| 2025-06-24 | Video-XL-2 | BAAI | model | 54 | |||||||||
| 2025-06-24 | Radial Attention | MIT | paper | ||||||||||
| 2025-06-23 | The Open Proof Corpus | ETH Zürich | dataset | ||||||||||
| 2025-06-18 | Show-o2 | NUS | model | 7B | |||||||||
| 2025-06-18 | Thunder-DeID | SNU | dataset | ||||||||||
| 2025-06-18 | ★ Thunder-LLM | SNU | model | ||||||||||
| 2025-06-18 | Thunder-Tok | SNU | paper | ||||||||||
| 2025-06-17 | ★ Mercury (Diffusion LLM) | Inception Labs | model | 1 | 128K | 14 (v4) | |||||||
| 2025-06-17 | PFMBench: Protein Foundation Model Benchmark | Westlake University | eval | 38 | |||||||||
| 2025-06-16 | SciSage / SurveyScope | BAAI | library | ||||||||||
| 2025-06-16 | ★ MiniMax-M1 2 | MiniMax | model paper | 3.15K | 855 | 456B | 45.9B | 1M | 12 (v4) | ||||
| 2025-06-15 | MOSS-TTSD | Fudan University | model | ||||||||||
| 2025-06-15 | ★ AI-Driven Agentic Design Platform for Tumor Immunotherapy Drugs | ZGCA | announcement | ||||||||||
| 2025-06-15 | PR[AI]RIE-PSAI AI Cluster Wins €75 Million in France 2030 Funding Université PSL | INRIA | news | ||||||||||
| 2025-06-15 | ZGCA & ZGCI Unveil AI-Driven Tumor Immunotherapy Drug Design Platform Zhongguancun Academy | ZGCA | news | ||||||||||
| 2025-06-14 | OpenUnlearning | CMU | library | ||||||||||
| 2025-06-13 | Scientists' First Exam | PJLab | eval | 830 | 66 | ||||||||
| 2025-06-12 | Seed-1.6 (AdaCoT) | ByteDance | model | 256K | |||||||||
| 2025-06-12 | ★ Magistral | Mistral | model | 38.49K | 2 | 24B (max) | 9 (v4) | ||||||
| 2025-06-12 | Predictable Scale Part II: Farseer | StepFun | paper | ||||||||||
| 2025-06-12 | Seed-1.6 Introduces Adaptive Chain-of-Thought (AdaCoT) ByteDance Seed | ByteDance | news | ||||||||||
| 2025-06-11 | FlagEvalMM | BAAI | library | 106 | |||||||||
| 2025-06-11 | Attention-Based Map Encoding (AME) | ETH Zürich | paper | ||||||||||
| 2025-06-10 | Seedance 1.0 | ByteDance | model | ||||||||||
| 2025-06-09 | RedNote joins AI race with its own open-source model that it says bests Alibaba, DeepSeek SCMP | Xiaohongshu | news | ||||||||||
| 2025-06-07 | ★ The Illusion of Thinking | Apple | paper | ||||||||||
| 2025-06-07 | Polaris | HKU | model | ||||||||||
| 2025-06-06 | RoboBrain 2.0 2 | BAAI | model | 1.09K | 573 | ||||||||
| 2025-06-06 | ★ MiniCPM4 2 | OpenBMB | model paper | 9.42K | 11.73K | 8B | |||||||
| 2025-06-06 | Ultra-FineWeb | OpenBMB | dataset | 30 | |||||||||
| 2025-06-06 | ★ dots.llm1 2 | Xiaohongshu | model paper | 142B | 14B | 11.2T | 32.77K | ||||||
| 2025-06-05 | RoboRefer / RefSpatial | BAAI | model | 263 | |||||||||
| 2025-06-05 | Boltzmann Sampler Line (PTSD / BNEM / DiKL) | University of Cambridge | paper | ||||||||||
| 2025-06-05 | Loss Deceleration & Zero-Sum Learning | Mila | paper | ||||||||||
| 2025-06-05 | Log-Linear Attention | MIT | paper | ||||||||||
| 2025-06-05 | The Common Pile 2 | Toronto & Vector Institute | dataset model | 7B | |||||||||
| 2025-06-04 | EuroLLM | University of Edinburgh | model | 22B (max) | |||||||||
| 2025-06-04 | MiMo-VL 2 | Xiaomi | model paper | 643 | 3.93K | 7B | |||||||
| 2025-06-04 | Chinese social media app Xiaohongshu's $26 billion valuation bolsters GSR fund Bloomberg | Xiaohongshu | news | ||||||||||
| 2025-06-03 | CyberGym | UC Berkeley | eval | ||||||||||
| 2025-06-03 | KRONOS | Harvard | model | ||||||||||
| 2025-06-03 | UniWorld | Peking University | model | ||||||||||
| 2025-06-02 | OWSM / OWLS | CMU | model | 18B | |||||||||
| 2025-06-01 | HumanSense Benchmark | Ant Group | dataset | ||||||||||
| 2025-06-01 | BrowseComp & WideSearch | Moonshot AI | dataset | ||||||||||
| 2025-06-01 | kimi-agent-sdk | Moonshot AI | library | 487 | |||||||||
| 2025-06-01 | kimi-cli | Moonshot AI | library | 8.94K | |||||||||
| 2025-06-01 | Kimi-Dev 2 | Moonshot AI | model paper | 1.23K | 2.67K | 72B | |||||||
| 2025-06-01 | Kimi-Researcher | Moonshot AI | model | 81 | |||||||||
| 2025-06-01 | walle | Moonshot AI | library | 21 | |||||||||
| 2025-06-01 | AgentCPM Series 3 | OpenBMB | paper | 800 | 1.67K | ||||||||
| 2025-06-01 | A.X Encoder | SK Telecom | model | 2.68K | |||||||||
| 2025-06-01 | CF-Div2-Stepfun | StepFun | dataset | ||||||||||
| 2025-06-01 | SteptronOss | StepFun | library | 575 | |||||||||
| 2025-06-01 | Cryo-IEF | Westlake University | model | ||||||||||
| 2025-06-01 | MiMo-Audio 2 | Xiaomi | model paper | 7B | |||||||||
| 2025-05-30 | AReaL | Ant Group | library | 5.29K | |||||||||
| 2025-05-30 | ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries | NVIDIA | paper | ||||||||||
| 2025-05-29 | GSO | UC Berkeley | eval | 102 | |||||||||
| 2025-05-29 | MathArena | ETH Zürich | eval | ||||||||||
| 2025-05-29 | KVzip | SNU | paper | ||||||||||
| 2025-05-29 | BioReason | Toronto & Vector Institute | model | ||||||||||
| 2025-05-29 | TrajViT | University of Washington | paper | ||||||||||
| 2025-05-28 | Ming-Omni | Ant Group | model | 656 | 42 | ||||||||
| 2025-05-28 | ★ DeepSeek-R1-0528 | DeepSeek | model | 671B | 37B | ||||||||
| 2025-05-28 | Pangu Embedded | Huawei | paper | 118 | |||||||||
| 2025-05-28 | ★ Skywork Open Reasoner 1 | Skywork | model | 743 | 32B | ||||||||
| 2025-05-27 | Pangu Pro MoE | Huawei | paper | 133 | 1 | ||||||||
| 2025-05-27 | AutoReproduce | OpenBMB | paper | ||||||||||
| 2025-05-27 | GTA & GLA — Hardware-Efficient Attention | Princeton | paper | ||||||||||
| 2025-05-27 | HunyuanVideo-Avatar | Tencent | paper | ||||||||||
| 2025-05-26 | FLAME-MoE | CMU | model | ||||||||||
| 2025-05-26 | SynLogic 2 | MiniMax | paper dataset | 203 | 121 | ||||||||
| 2025-05-26 | VLM-3R | UT Austin | paper | ||||||||||
| 2025-05-26 | VoiceStar | UT Austin | model | ||||||||||
| 2025-05-25 | ★ FP4 All the Way | Technion | paper | ||||||||||
| 2025-05-24 | AI-Researcher | HKU | library | ||||||||||
| 2025-05-23 | Why Diffusion Models Don't Memorize | INRIA | paper | ||||||||||
| 2025-05-23 | One RL to See Them All: Visual Triple Unified RL | MiniMax | paper | 333 | |||||||||
| 2025-05-23 | FairyR1-32B | Peking University | model | ||||||||||
| 2025-05-22 | ★ Claude 4 | Anthropic | model | ~1.4T | 1M | 21 (v4) | |||||||
| 2025-05-22 | The Polar Express | NYU | paper | ||||||||||
| 2025-05-22 | XRing O1 | Xiaomi | announcement | ||||||||||
| 2025-05-21 | Reverse Engineering Human Preferences with RL | Imperial | paper | ||||||||||
| 2025-05-21 | Devstral 2 | Mistral | model | 123B | 256K | 9 (v4) | |||||||
| 2025-05-21 | ★ Falcon-H1 | TII | model | 118 | 11.02K | 34B | 256K | ||||||
| 2025-05-20 | BAGEL | ByteDance | model | 6K | 820 | 14B | |||||||
| 2025-05-20 | s3 | UIUC | paper | ||||||||||
| 2025-05-19 | On the Conversational Persuasiveness of GPT-4 | EPFL | paper | ||||||||||
| 2025-05-19 | ★ Marin | Open Athena, Stanford | model | 32B | |||||||||
| 2025-05-19 | ★ Terminal-Bench | Stanford | eval | 66 | |||||||||
| 2025-05-19 | Quetzal | Toronto & Vector Institute | model | ||||||||||
| 2025-05-18 | mCLM | UIUC | paper | ||||||||||
| 2025-05-17 | Video-SafetyBench | BAAI | eval | 12 | 2.26K | ||||||||
| 2025-05-17 | Model Merging in Pre-training of LLMs | ByteDance | paper | ||||||||||
| 2025-05-15 | BGE-Code-v1 | BAAI | model | 11.8K | 4.84K | ||||||||
| 2025-05-15 | MMLongBench | University of Edinburgh | eval | 13.33K | |||||||||
| 2025-05-15 | Superposition Yields Robust Neural Scaling | MIT | paper | ||||||||||
| 2025-05-15 | Apriel-Nemotron-15B Reasoning Model with NVIDIA ServiceNow | ServiceNow | news | ||||||||||
| 2025-05-14 | ★ AlphaEvolve | paper | 6 | ||||||||||
| 2025-05-13 | HealthBench 2 | OpenAI | eval | 5K | |||||||||
| 2025-05-12 | Seed1.5-VL | ByteDance | model | 1.58K | 200B | 20B | |||||||
| 2025-05-12 | MiniMax-Speech: Intrinsic Zero-Shot TTS | MiniMax | paper | ||||||||||
| 2025-05-12 | Step1X-3D: High-Fidelity Textured 3D Assets | StepFun | model | 870 | |||||||||
| 2025-05-11 | GuidedQuant | SNU | paper | ||||||||||
| 2025-05-10 | Gated Attention for Large Language Models | Alibaba | paper | ||||||||||
| 2025-05-08 | Seed-Coder-8B | ByteDance | model | 755 | 8B | 65.54K | |||||||
| 2025-05-07 | DeerFlow | ByteDance | library | 70.89K | |||||||||
| 2025-05-07 | ★ Pangu Ultra MoE 2 | Huawei | model paper | 16 | 718B | 39B | |||||||
| 2025-05-07 | General-Level & General-Bench | NUS | eval | 700 | |||||||||
| 2025-05-07 | HunyuanCustom | Tencent | paper | 1.22K | 1 | ||||||||
| 2025-05-06 | CCI 4.0 | BAAI | dataset | ||||||||||
| 2025-05-06 | OpenSeek | BAAI | model | 261 | 4 | ||||||||
| 2025-05-06 | RoboOS 2 | BAAI | library | 575 | |||||||||
| 2025-05-06 | SkyRL | UC Berkeley | library | ||||||||||
| 2025-05-06 | VideoMimic | UC Berkeley | paper | ||||||||||
| 2025-05-06 | WebGen-Bench | CUHK | eval | 647 | 101 | ||||||||
| 2025-05-06 | Absolute Zero Reasoner | Tsinghua | paper | ||||||||||
| 2025-05-06 | OpenHelix | Westlake University | model | ||||||||||
| 2025-05-05 | El Agente Q | Toronto & Vector Institute | paper | ||||||||||
| 2025-05-02 | ★ MiMo (Reasoning) 2 | Xiaomi | model paper | 91.71K | 1 | 7B | |||||||
| 2025-05-01 | AWorld | Ant Group | library | 1.2K | |||||||||
| 2025-05-01 | ★ Kanana 1.5 | Kakao | model | 89 | 15.7B | 3B | 32K | ||||||
| 2025-05-01 | M2PDE / RealPDEBench | Westlake University | paper | ||||||||||
| 2025-04-30 | ★ Amazon Nova Premier | Amazon | model | ~470B | 1M | 9 (v4) | |||||||
| 2025-04-30 | DeepSeek-Prover-V2 | DeepSeek | model | 1.27K | 633 | 671B | |||||||
| 2025-04-30 | SWE-smith | Princeton | dataset | ||||||||||
| 2025-04-30 | WebThinker | Renmin (RUC) | paper | ||||||||||
| 2025-04-30 | Nova Premier Launched as Amazon's Most Capable AI Model TechCrunch | Amazon | news | ||||||||||
| 2025-04-30 | Phi-4 Reasoning Models Released with Chain-of-Thought Microsoft | Microsoft | news | ||||||||||
| 2025-04-29 | ★ Qwen3 9 | Alibaba | model paper | 27.3K | 74 | 1T | 22B (max) | 9 (v4) | |||||
| 2025-04-29 | Behavior-SD | SNU | dataset | ||||||||||
| 2025-04-29 | First LlamaCon Developer Conference Meta AI | Meta | news | ||||||||||
| 2025-04-27 | Explanatory Summarization with Discourse-Driven Planning | University of Edinburgh | paper | ||||||||||
| 2025-04-25 | PolyMath | Alibaba | dataset | 44 | |||||||||
| 2025-04-25 | Kimi-Audio 2 | Moonshot AI | model paper | 4.65K | 80.22K | 7B | |||||||
| 2025-04-24 | gRNAde | University of Cambridge | model | ||||||||||
| 2025-04-24 | The Sparse Frontier | University of Edinburgh | paper | ||||||||||
| 2025-04-24 | Paper2Code | KAIST | library | ||||||||||
| 2025-04-24 | Step1X-Edit | StepFun | model | 2.22K | 110 | ||||||||
| 2025-04-22 | LiveCC | NUS | model | ||||||||||
| 2025-04-22 | TTRL: Test-Time Reinforcement Learning | PJLab | paper | ||||||||||
| 2025-04-21 | LUFFY: Learning to Reason under Off-Policy Guidance | Westlake University | paper | ||||||||||
| 2025-04-19 | SRPO: Staged History-Resampling Policy Optimization | Kuaishou | paper | ||||||||||
| 2025-04-18 | Does RL Really Incentivize Reasoning Beyond the Base Model? | Tsinghua | paper | ||||||||||
| 2025-04-17 | Nemotron-CLIMB: Clustering-based Iterative Data Mixture Bootstrapping 3 | NVIDIA | paper dataset | 1 | |||||||||
| 2025-04-16 | BrowseComp | OpenAI | eval | 1.27K | |||||||||
| 2025-04-16 | ★ o3 | OpenAI | model | ~3T | 200K | 20 (v4) | |||||||
| 2025-04-15 | DataDecide | Ai2 | paper | ||||||||||
| 2025-04-15 | ReTool: Reinforcement Learning for Strategic Tool Use in LLMs | ByteDance | paper | 2 | |||||||||
| 2025-04-15 | ★ Kling 2.0 1 | Kuaishou | model | ||||||||||
| 2025-04-15 | Kimina-Prover 2 | Moonshot AI | model paper | 370 | 717 | ||||||||
| 2025-04-15 | miniF2F-test (Rectified) | Moonshot AI | dataset | 370 | |||||||||
| 2025-04-15 | ★ Apriel | ServiceNow | model | 3 | 24 | 15B (max) | 4.5T | ||||||
| 2025-04-15 | Step-R1-V-Mini | StepFun | model | ||||||||||
| 2025-04-15 | ZR1-1.5B | Zyphra | model | 1.5K | 1.5B | ||||||||
| 2025-04-15 | Apriel-5B: ServiceNow's First Open SLM ServiceNow | ServiceNow | news | ||||||||||
| 2025-04-14 | SocioVerse | Fudan University | library | ||||||||||
| 2025-04-14 | InternVL3 | PJLab | model | 10.06K | 5 | 78B (max) | |||||||
| 2025-04-12 | SenseNova V6 | SenseTime | model | 600B | |||||||||
| 2025-04-11 | CellFlow | TU Munich | paper | ||||||||||
| 2025-04-10 | ★ Scaling Laws for Native Multimodal Models | Apple | paper | ||||||||||
| 2025-04-10 | ★ Seed1.5-Thinking: Advancing Superb Reasoning Models with RL | ByteDance | paper | 1 | |||||||||
| 2025-04-10 | ★ Pangu Ultra 2 | Huawei | model paper | 77 | 135B | ||||||||
| 2025-04-10 | Kimi-VL 2 | Moonshot AI | model paper | 1.2K | 123.09K | 1 | 16B | 2.8B | |||||
| 2025-04-09 | SoK: Membership Inference Attacks on LLMs | Imperial | paper | ||||||||||
| 2025-04-09 | DisCIPL (Self-Steering LMs) | MIT | paper | ||||||||||
| 2025-04-09 | A Sober Look at Progress in LM Reasoning | Tübingen | paper | ||||||||||
| 2025-04-08 | Amazon Nova Sonic | Amazon | model | ||||||||||
| 2025-04-08 | Dream 7B | HKU, Huawei | model | 1.25K | 7B | ||||||||
| 2025-04-08 | ★ Skywork R1V Series | Skywork | model | 41 | 38B | ||||||||
| 2025-04-08 | Nova Sonic Speech-to-Speech Model Launched on Bedrock AWS | Amazon | news | ||||||||||
| 2025-04-07 | BaichuanMed-OCR | Baichuan | model | 169 | 72B (max) | ||||||||
| 2025-04-07 | EvoTune | EPFL | paper | ||||||||||
| 2025-04-06 | Metamon | UT Austin | paper | ||||||||||
| 2025-04-05 | ★ Llama 4 | Meta | model | 400B | 17B | 1M | 9 (v4) | ||||||
| 2025-04-05 | Llama 4 Scout and Maverick Released (First MoE, Multimodal) Meta AI | Meta | news | ||||||||||
| 2025-04-04 | ★ Nemotron-H | NVIDIA | model | 58.35K | 56B | 20T | |||||||
| 2025-04-04 | DeepResearcher | Shanghai Jiao Tong University | model | 7B | |||||||||
| 2025-04-04 | MedSAM2 | Toronto & Vector Institute | model | ||||||||||
| 2025-04-03 | DeepSeek-GRM: Inference-Time Scaling for Generalist Reward Modeling | DeepSeek | paper | 1 | |||||||||
| 2025-04-02 | ATOMICA | Harvard | model | ||||||||||
| 2025-04-01 | MiniMax Speech Series | MiniMax | model | ||||||||||
| 2025-03-31 | Amazon Nova Act | Amazon | model | 909 | |||||||||
| 2025-03-31 | Open-Reasoner-Zero: Scaling Up RL on the Base Model | StepFun, Tsinghua | paper | ||||||||||
| 2025-03-31 | Amazon Unveils Nova Act, an AI Agent That Controls a Web Browser TechCrunch | Amazon | news | ||||||||||
| 2025-03-30 | ★ ToRL: Scaling Tool-Integrated RL | SII | paper | 349 | 1 | ||||||||
| 2025-03-28 | Doubao-Deep-Thinking | ByteDance | model | ||||||||||
| 2025-03-27 | OpenComplex 2 | BAAI | model | 269 | |||||||||
| 2025-03-27 | OlymMATH | Renmin (RUC) | eval | ||||||||||
| 2025-03-26 | Qwen2.5-Omni-7B | Alibaba | model | 4.02K | 777.76K | 7B | |||||||
| 2025-03-25 | ★ Gemini 2.5 Pro | model | ~1.2T | 1M | 16 (v4) | ||||||||
| 2025-03-24 | SimpleRL-Zoo: Investigating and Taming Zero RL for Open Base Models | ByteDance, Meituan | paper | ||||||||||
| 2025-03-24 | CaMeL | ETH Zürich | paper | ||||||||||
| 2025-03-24 | ★ SimpleRL / SimpleRL-Zoo | HKUST | model | ||||||||||
| 2025-03-22 | Safe RLHF-V | Peking University | paper | ||||||||||
| 2025-03-21 | Hunyuan-T1 | Tencent | model | ||||||||||
| 2025-03-21 | CVE-Bench | UIUC | eval | ||||||||||
| 2025-03-21 | SEA-HELM: Assessing LLM Performance for Southeast Asia AI Singapore | AI Singapore | news | ||||||||||
| 2025-03-20 | ★ Aardvark Weather | University of Cambridge | model | ||||||||||
| 2025-03-18 | Sable | BAAI | model | 2 | |||||||||
| 2025-03-18 | Isaac GR00T | NVIDIA | model | ||||||||||
| 2025-03-18 | ★ Llama-Nemotron (Nano/Super/Ultra) | NVIDIA | model | 123.53K | 1 | 128K | 8 (v4) | ||||||
| 2025-03-18 | 1000-Layer Networks for Self-Supervised RL | Princeton | paper | ||||||||||
| 2025-03-18 | HaploVL | Tencent | model | 65 | |||||||||
| 2025-03-17 | DAPO 2 | ByteDance | paper dataset | ||||||||||
| 2025-03-17 | ★ EXAONE Deep | LG | model | 1.4K | 32B | ||||||||
| 2025-03-17 | SuperBPE | University of Washington | paper | ||||||||||
| 2025-03-16 | ★ ERNIE 4.5 | Baidu | model | 7.72K | 424B (max) | 47B (max) | 8 (v4) | ||||||
| 2025-03-16 | ERNIE X1 | Baidu | model | ||||||||||
| 2025-03-14 | TxAgent & ToolUniverse | Harvard | library | ||||||||||
| 2025-03-14 | VGGT | University of Oxford, Meta | model | ||||||||||
| 2025-03-13 | MMLU-ProX | UTokyo | eval | ||||||||||
| 2025-03-13 | BOSS直聘发布全年财报:2024年营收73.56亿元,同比上涨23.6% (BOSS Zhipin full-year results; Nanbeige plus DeepSeek-R1 in the app) Xinhua | Nanbeige | news | ||||||||||
| 2025-03-12 | ★ Gemma 3 | model | 1.42M | 40 | 27B | 128K | 5 (v4) | ||||||
| 2025-03-12 | Search-R1 | UIUC | paper | ||||||||||
| 2025-03-11 | TinyDeepSeek | CUHK | model | 3.3B | |||||||||
| 2025-03-10 | Seedream 2.0 | ByteDance | paper | ||||||||||
| 2025-03-10 | YOLOE | Tsinghua | model | ||||||||||
| 2025-03-10 | Reka Flash 3 Released (Open-Weight, 21B) Reka | Reka | news | ||||||||||
| 2025-03-07 | ★ Ling 2 | Ant Group | paper model | 258 | 38.63K | 1 | 16B | ||||||
| 2025-03-07 | R1-Searcher | Renmin (RUC) | paper | ||||||||||
| 2025-03-06 | QwQ-32B | Alibaba | model | 52.45K | 32B | 9 (v4) | |||||||
| 2025-03-06 | BGE-VL 2 | BAAI | model dataset | 11.8K | 4.55K | ||||||||
| 2025-03-06 | L1 (Length Controlled Policy Optimization) | CMU | paper | ||||||||||
| 2025-03-06 | Predictable Scale Part I: Step Law | StepFun | paper | 1 | |||||||||
| 2025-03-05 | EgoLife & EgoGPT | NTU | model | ||||||||||
| 2025-03-05 | DINGO-BNS | Tübingen | paper | ||||||||||
| 2025-03-04 | CogView-4 | Z.ai | model | 1.1K | |||||||||
| 2025-03-03 | ★ Aya Vision | Cohere | model | 160.27K | 32B | ||||||||
| 2025-03-01 | ★ Command A | Cohere | model | 2.28K | 111B | 256K | 7 (v4) | ||||||
| 2025-03-01 | ★ Sarashina2.2 3 | SB Intuitions | model | 12.98K | 4B | ||||||||
| 2025-02-28 | 3FS (Fire-Flyer File System) | DeepSeek | library | 9.96K | |||||||||
| 2025-02-28 | Smallpond | DeepSeek | library | 4.96K | |||||||||
| 2025-02-28 | Image-01 | MiniMax | model | ||||||||||
| 2025-02-27 | RoboBrain | BAAI | model | 553 | 132 | ||||||||
| 2025-02-27 | UniTok | ByteDance | paper | 526 | |||||||||
| 2025-02-27 | DualPipe | DeepSeek | library | 2.96K | |||||||||
| 2025-02-27 | EPLB (Expert Parallelism Load Balancer) | DeepSeek | library | 1.39K | |||||||||
| 2025-02-27 | Hunyuan Turbo S 2 | Tencent | model paper | 560B | 56B | ||||||||
| 2025-02-27 | moscot | TU Munich | library | ||||||||||
| 2025-02-26 | DeepGEMM | DeepSeek | library | 7.36K | |||||||||
| 2025-02-26 | BIG-Bench Extra Hard (BBEH) | eval | 4.52K | 23 | |||||||||
| 2025-02-26 | ★ Kanana | Kakao | model | 280 | 1 | 32.5B | 3T | ||||||
| 2025-02-26 | NeoBERT | Mila | model | 250M | |||||||||
| 2025-02-26 | Granite 3.2: Multimodal Vision and Chain-of-Thought Reasoning IBM | IBM | news | ||||||||||
| 2025-02-25 | DeepEP | DeepSeek | library | 9.71K | |||||||||
| 2025-02-25 | Rank1 | Johns Hopkins | model | ||||||||||
| 2025-02-24 | ★ Claude Code | Anthropic | library | 131.53K | |||||||||
| 2025-02-24 | Baichuan-Audio 2 | Baichuan | paper model | 222 | 103 | 10B | |||||||
| 2025-02-24 | FlashMLA | DeepSeek | library | 12.7K | |||||||||
| 2025-02-24 | Reasoning with Latent Thoughts: On the Power of Looped Transformers | paper | 1 | ||||||||||
| 2025-02-24 | Muon Optimizer 2 | Moonshot AI | paper library | 1 | |||||||||
| 2025-02-24 | Topic Over Source: The Key to Effective Data Mixing for LLM Pre-training | PJLab | paper | 1 | |||||||||
| 2025-02-24 | AISafetyLab | Tsinghua | library | ||||||||||
| 2025-02-24 | Erwin | University of Amsterdam | paper | ||||||||||
| 2025-02-22 | Moonlight-3B/16B | Moonshot AI | model | 1.49K | 32.57K | 1 | 16B (max) | ||||||
| 2025-02-21 | LightThinker | Zhejiang University | paper | ||||||||||
| 2025-02-20 | SEA-HELM | AI Singapore | eval | ||||||||||
| 2025-02-20 | ★ SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines | ByteDance | eval | 188 | 4 | 26.53K | 285 | ||||||
| 2025-02-20 | Measuring CoT Faithfulness by Unlearning Reasoning Steps | Technion | paper | ||||||||||
| 2025-02-19 | Qwen2.5-VL | Alibaba | model | 52 | |||||||||
| 2025-02-19 | FlexTok | Apple | paper | 319 | |||||||||
| 2025-02-19 | MCQA critique: A) Forced B) Flawed C) Fixable D) All of the Above | University of Maryland | paper | ||||||||||
| 2025-02-18 | MoBA: Mixture of Block Attention for Long-Context LLMs | Moonshot AI | paper | 2.13K | 2 | ||||||||
| 2025-02-18 | Hunyuan-Large-Vision | Tencent | model | 389B | 52B | ||||||||
| 2025-02-18 | CuatroLLM / TransWebLLM | University College London | model | 1.3B | |||||||||
| 2025-02-17 | Mistral Saba | Mistral | model | 24B | 32K | 6 (v4) | |||||||
| 2025-02-17 | OpenDWM / MaskGWM 2 | SenseTime | library paper | 398 | |||||||||
| 2025-02-17 | ★ Grok-3 | SpaceXAI | model | ~2.1T | 1M | 12 (v4) | |||||||
| 2025-02-17 | Step-Audio / Step-Audio2 | StepFun | model | 27 | |||||||||
| 2025-02-17 | Grok-3 Launched, Trained on 200K GPU Colossus Cluster xAI | SpaceXAI | news | ||||||||||
| 2025-02-16 | AdaGC: Improving Training Stability for Large Language Model Pretraining | Baidu | paper | ||||||||||
| 2025-02-16 | NSA: Native Sparse Attention | DeepSeek | paper | 2 | |||||||||
| 2025-02-15 | 1bit-Merging 1 | Huawei | paper | ||||||||||
| 2025-02-14 | WebOrganizer | Ai2 | paper | ||||||||||
| 2025-02-14 | ★ LLaDA 2 | Ant Group | model | 3.82K | 2.27K | 5 | |||||||
| 2025-02-14 | FineWeb2-HQ / FineWeb-HQ | EPFL | dataset | ||||||||||
| 2025-02-14 | KernelBench | Stanford | eval | ||||||||||
| 2025-02-14 | Step-Video-T2V 2 | StepFun | model paper | 3.19K | 300B | ||||||||
| 2025-02-13 | LOBS5 / LOB-Bench | University of Oxford | model | ||||||||||
| 2025-02-12 | WorldGUI & GUI-Thinker | NUS | eval | ||||||||||
| 2025-02-12 | AgentSociety | Tsinghua | library | ||||||||||
| 2025-02-12 | Wu Yonghui Joins ByteDance as Head of Seed Basic Research SCMP | ByteDance | news | ||||||||||
| 2025-02-11 | Scion | EPFL | paper | ||||||||||
| 2025-02-11 | CodeI/O | HKUST | dataset | ||||||||||
| 2025-02-11 | ★ Nature Language Model (NatureLM) | Microsoft | model | 46 | 1 | 46.7B | 13B | ||||||
| 2025-02-11 | TransMLA | Peking University | paper | ||||||||||
| 2025-02-11 | Goedel-Prover | Princeton | model | ||||||||||
| 2025-02-10 | Agentica RL Line (DeepScaleR / DeepCoder / DeepSWE) | UC Berkeley | model | ||||||||||
| 2025-02-10 | QServe / LServe (OmniServe) | MIT | library | ||||||||||
| 2025-02-09 | AutoAgent | HKU | library | ||||||||||
| 2025-02-08 | TabICL / CARTE | INRIA | model | ||||||||||
| 2025-02-07 | ITBench | IBM | eval | 94 | |||||||||
| 2025-02-07 | ★ Huginn-3.5B | Tübingen, University of Maryland | model | 3.5B | 800B | ||||||||
| 2025-02-07 | ★ Gemstones | University of Maryland | model | 2B | |||||||||
| 2025-02-06 | Great Models Think Alike | Tübingen | paper | ||||||||||
| 2025-02-05 | ★ Scaling Laws for Upcycling Mixture-of-Experts Language Models | SB Intuitions | paper | 9 | |||||||||
| 2025-02-05 | ★ LIMO | SII | paper | 54 | 3 | ||||||||
| 2025-02-04 | OpenAI and Kakao to Jointly Develop AI Products for South Korea CNBC | Kakao | news | ||||||||||
| 2025-02-03 | TwinMarket | CUHK | library | ||||||||||
| 2025-02-02 | Sundial / Timer-XL | Tsinghua | model | ||||||||||
| 2025-02-01 | ModernBERT-Ja | SB Intuitions | model | 44.81K | 310M | ||||||||
| 2025-01-31 | LR Schedules and Convex Optimization | INRIA | paper | ||||||||||
| 2025-01-31 | s1: Simple Test-Time Scaling | Stanford | paper | ||||||||||
| 2025-01-30 | MedXpertQA: Benchmarking Expert-Level Medical Reasoning and Understanding | PJLab | paper | ||||||||||
| 2025-01-30 | MedXpertQA | PJLab | eval | 4.46K | |||||||||
| 2025-01-29 | International AI Safety Report 2 | Mila | paper | ||||||||||
| 2025-01-28 | Over-Tokenized Transformer | ByteDance | paper | ||||||||||
| 2025-01-28 | THREADS | Harvard | model | ||||||||||
| 2025-01-27 | Training Dynamics of In-Context Learning in Linear Attention | University College London | paper | ||||||||||
| 2025-01-26 | Baichuan-Omni-1.5 2 | Baichuan | paper model | 191 | 1.84K | 7B | |||||||
| 2025-01-24 | Baichuan-M1 3 | Baichuan | paper model | 821 | 5 | 14.5B | |||||||
| 2025-01-24 | FireRedASR | Xiaohongshu | model | 8.3B (max) | |||||||||
| 2025-01-23 | Video-MMMU | NTU | eval | 900 | |||||||||
| 2025-01-23 | UltraRAG | OpenBMB | library | 5.58K | |||||||||
| 2025-01-22 | ★ Doubao-1.5-Pro | ByteDance | model | 200B | 20B | 256K | |||||||
| 2025-01-22 | UI-TARS | ByteDance | library | 10.9K | 621.09K | 5 | |||||||
| 2025-01-22 | ★ DeepSeek-R1 | DeepSeek | model | 5.35M | 671B | 37B | 13 (v4) | ||||||
| 2025-01-22 | Revisit Self-Debugging with Self-Generated Tests for Code Generation | Meituan | paper | 1 | |||||||||
| 2025-01-21 | TokenVerse | Technion | paper | ||||||||||
| 2025-01-21 | Hunyuan3D 2.0 3 | Tencent | model paper | 13.93K | 75.95K | 6 | |||||||
| 2025-01-21 | Stargate Project: $500B AI Infrastructure Initiative OpenAI | OpenAI | news | ||||||||||
| 2025-01-20 | ★ Kimi k1.5 2 | Moonshot AI | model paper | 3.47K | 11 | ||||||||
| 2025-01-17 | ComplexFuncBench | Z.ai | dataset | 180 | |||||||||
| 2025-01-15 | ★ SpeechGPT-2.0-preview | Fudan University | model | 7B | |||||||||
| 2025-01-15 | UNI2-h & CONCH v1.5 | Harvard | model | ||||||||||
| 2025-01-15 | ★ InternLM3 | PJLab | model | 7.22K | |||||||||
| 2025-01-14 | ★ MiniMax-01 3 | MiniMax | model paper | 3.43K | 172.55K | 1 | 456B | 4M | |||||
| 2025-01-14 | ★ MiniCPM-o 2.6 | OpenBMB | model | 25.58K | 386.43K | 8B | |||||||
| 2025-01-14 | InstructCell | Zhejiang University | model | ||||||||||
| 2025-01-10 | ★ Sky-T1 / SkyThought | UC Berkeley | model | ||||||||||
| 2025-01-10 | GThinker | PCL | model | ||||||||||
| 2025-01-09 | Search-o1 | Renmin (RUC) | paper | ||||||||||
| 2025-01-09 | WanJuan 3.0 (WanJuan-SiLu) | PJLab | dataset | ||||||||||
| 2025-01-09 | Asagi | UTokyo | model | ||||||||||
| 2025-01-07 | ★ Cosmos | NVIDIA | model | 9 | |||||||||
| 2025-01-06 | Evolla | Westlake University | model | 10B (max) | |||||||||
| 2025-01-03 | AgentRefine | Meituan | paper | ||||||||||
| 2025-01-02 | SEDD (GPU deduplication) | SNU | paper | ||||||||||
| 2025-01-02 | FlashInfer | University of Washington | library | ||||||||||
| 2025-01-01 | Document Parse | Upstage | library | ||||||||||
| 2024-12-31 | ★ OLMo 2 | Ai2 | model | 3.72K | 2 | 32B | 6T | 4.1K | 6 (v4) | ||||
| 2024-12-30 | SWE-Gym | UC Berkeley | dataset | ||||||||||
| 2024-12-26 | ★ DeepSeek-V3 3 | DeepSeek | model paper | 227 | 671B | 37B | 8 (v4) | ||||||
| 2024-12-25 | QVQ | Alibaba | model | ||||||||||
| 2024-12-25 | ★ HuatuoGPT-o1 | CUHK | model | ||||||||||
| 2024-12-24 | ★ LLM-jp-3 (172B) | NII | model | 20 | 172B | 2.1T | 4.1K | ||||||
| 2024-12-23 | Baichuan4-Finance 2 | Baichuan | paper model | 1 | |||||||||
| 2024-12-20 | Align-Anything | Peking University | library | ||||||||||
| 2024-12-19 | SLAM-LLM | Shanghai Jiao Tong University | library | ||||||||||
| 2024-12-19 | DreMa | University of Amsterdam | paper | ||||||||||
| 2024-12-18 | NOVA (Non-quantized Video Autoregressive) | BAAI | model | 651 | |||||||||
| 2024-12-18 | TheAgentCompany | CMU | eval | ||||||||||
| 2024-12-18 | VSI-Bench | NYU | eval | ||||||||||
| 2024-12-16 | ★ MASt3R-SLAM | Imperial | library | ||||||||||
| 2024-12-13 | DeepSeek-VL2 | DeepSeek | model | 5.3K | 2.68K | 22 | |||||||
| 2024-12-13 | Liquid AI Raises $250M Series A Led by AMD Liquid AI | Liquid AI | news | ||||||||||
| 2024-12-13 | Profile: Shanghai AI Lab: Driving both AI safety and development MERICS | PJLab | news | ||||||||||
| 2024-12-12 | ★ Phi-4 | Microsoft | model | 801.85K | 24 | 14B | 6 (v4) | ||||||
| 2024-12-12 | Phi-4 Released: 14B SLM Specializing in Complex Reasoning Microsoft Research | Microsoft | news | ||||||||||
| 2024-12-11 | SEA-LION v3 | AI Singapore | model | 70B | |||||||||
| 2024-12-11 | FlowEdit | Technion | paper | ||||||||||
| 2024-12-10 | ProCyon | Harvard | model | ||||||||||
| 2024-12-09 | ProcessBench | Alibaba | dataset | 1 | |||||||||
| 2024-12-07 | SNU Korean Eval Suite | SNU | eval | ||||||||||
| 2024-12-06 | ★ Aya Expanse | Cohere | model | 48.1K | 3 | 32B | 128K | ||||||
| 2024-12-06 | ★ EXAONE 3.5 | LG | model | 10.25K | 2 | 32B | |||||||
| 2024-12-06 | Densing Law of LLMs | OpenBMB | paper | 2 | |||||||||
| 2024-12-06 | InternVL 2.5 | PJLab | model | 10.06K | 349 | 13 | 78B (max) | ||||||
| 2024-12-06 | APOLLO | UT Austin | library | ||||||||||
| 2024-12-05 | Language Model Ladders | Ai2 | paper | ||||||||||
| 2024-12-05 | Infinity & InfinityStar | ByteDance | model | 1.57K | 1 | ||||||||
| 2024-12-05 | Liquid: Scalable Multi-modal Generation | ByteDance | model | 643 | 1 | ||||||||
| 2024-12-05 | Divot | Tencent | model | 87 | |||||||||
| 2024-12-05 | Moto | Tencent | paper | 177 | |||||||||
| 2024-12-04 | AceGPT & Native Alignment | CUHK | model | ||||||||||
| 2024-12-04 | GenCast | paper | |||||||||||
| 2024-12-04 | RedStone | Microsoft | dataset | 161 | |||||||||
| 2024-12-03 | Amazon Nova | Amazon | model | 2 | ~90B | 300K | 7 (v4) | ||||||
| 2024-12-03 | AWS Trainium2 (Trn2 / Trn2 UltraServer) | Amazon | announcement | ||||||||||
| 2024-12-03 | HunyuanVideo | Tencent | model | 12.19K | 6 | 13B | |||||||
| 2024-12-03 | SEED-Voken | Tencent | paper | 1.01K | |||||||||
| 2024-12-03 | GLM-4-Voice: End-to-End Spoken Chatbot | Z.ai | model | 2 | |||||||||
| 2024-12-03 | AWS Trainium2 Chips Generally Available; Trainium3 Previewed TechCrunch | Amazon | news | ||||||||||
| 2024-12-01 | ★ flash-linear-attention | MIT | library | ||||||||||
| 2024-12-01 | ★ Falcon 3 | TII | model | 10B | |||||||||
| 2024-12-01 | ScanNet++ (v2) | TU Munich | dataset | ||||||||||
| 2024-12-01 | Rainbow Teaming | University College London, Meta | paper | ||||||||||
| 2024-11-29 | ★ TITAN | Harvard | model | ||||||||||
| 2024-11-28 | ★ Open-Sora Plan | Peking University | model | 8B | |||||||||
| 2024-11-27 | Spatiotemporal Skip Guidance | KAIST | paper | ||||||||||
| 2024-11-26 | ShowUI | NUS | model | ||||||||||
| 2024-11-25 | Model Context Protocol (MCP) | Anthropic | library | ||||||||||
| 2024-11-22 | ★ Tülu 3 | Ai2 | model | 3.75K | 2.6K | 7 | |||||||
| 2024-11-22 | XGrammar | CMU | library | ||||||||||
| 2024-11-22 | ★ Zamba2 (Hybrid SSM/Transformer Suite) | Zyphra | model | 193 | 110 | 3 | 7.4B | 3T | |||||
| 2024-11-22 | Amazon Doubles Anthropic Investment to $8 Billion CNBC | Amazon | news | ||||||||||
| 2024-11-21 | ★ AIMv2 | Apple | paper | 1.42K | 2 | ||||||||
| 2024-11-21 | DINO-X 2 | IDEA Lab | model paper | 1.39K | 4 | ||||||||
| 2024-11-21 | Natural Language Reinforcement Learning | University College London | paper | ||||||||||
| 2024-11-20 | Hymba | NVIDIA | paper | 279 | 2 | ||||||||
| 2024-11-20 | BALROG | University College London | eval | ||||||||||
| 2024-11-19 | Aquila-VL-2B | BAAI | model | 49 | |||||||||
| 2024-11-19 | Loss-to-Loss Prediction | Harvard | paper | ||||||||||
| 2024-11-18 | Pixtral Large | Mistral | model | 124B | 128K | 7 (v4) | |||||||
| 2024-11-18 | STILL (Slow Thinking with LLMs) | Renmin (RUC) | paper | ||||||||||
| 2024-11-15 | MARS (Make vAriance Reduction Shine) | ByteDance | paper | 721 | |||||||||
| 2024-11-15 | LLaVA-CoT | Peking University | model | ||||||||||
| 2024-11-13 | CamemBERT | INRIA | model | 112M | |||||||||
| 2024-11-12 | Spider 2.0 | HKU | eval | 632 | |||||||||
| 2024-11-08 | ★ Sarashina2-8x70B | SB Intuitions | model | 12 | |||||||||
| 2024-11-08 | SB Intuitions releases 460B-parameter Japanese LLM Sarashina2-8x70B for academia and industry SB Intuitions | SB Intuitions | news | ||||||||||
| 2024-11-07 | SVDQuant + Nunchaku | MIT | paper | ||||||||||
| 2024-11-07 | OneProt | TU Munich | paper | ||||||||||
| 2024-11-06 | Touchstone | Johns Hopkins | eval | ||||||||||
| 2024-11-04 | ★ Hunyuan-Large 2 | Tencent | model paper | 1.59K | 482 | 6 | 389B | 52B | 7T | 256K | |||
| 2024-11-04 | Hunyuan3D 1.0 | Tencent | model | 3.48K | |||||||||
| 2024-11-04 | TableGPT2 | Zhejiang University | model | 72B | |||||||||
| 2024-11-01 | SimpleQA | OpenAI | eval | 12 | 4.33K | ||||||||
| 2024-11-01 | InternThinker | PJLab | model | ||||||||||
| 2024-11-01 | TREC iKAT | University of Amsterdam | eval | ||||||||||
| 2024-10-31 | SelfCodeAlign | UIUC | paper | ||||||||||
| 2024-10-30 | Kinetix | University of Oxford | paper | ||||||||||
| 2024-10-29 | Agentforce Platform Launched for Enterprise AI Agents Salesforce | Salesforce | news | ||||||||||
| 2024-10-28 | Arithmetic Without Algorithms | Technion | paper | ||||||||||
| 2024-10-28 | AutoGLM 2 | Z.ai | model paper | 1 | |||||||||
| 2024-10-28 | Zhongguancun Institute of Artificial Intelligence Established ZGCI | ZGCA | news | ||||||||||
| 2024-10-27 | Llama Scope | Fudan University | paper | ||||||||||
| 2024-10-27 | ThunderKittens | Stanford | library | ||||||||||
| 2024-10-25 | SaprotHub | Westlake University | library | ||||||||||
| 2024-10-24 | Infinity-MM | BAAI | dataset | ||||||||||
| 2024-10-24 | MotionCLR 2 | IDEA Lab | library paper | 17 | |||||||||
| 2024-10-24 | O1 Replication Journey | Shanghai Jiao Tong University | paper | ||||||||||
| 2024-10-24 | ★ Skywork-Reward 2 | Skywork | model | 37 | |||||||||
| 2024-10-23 | DiffuGPT / DiffuLLaMA | HKU | paper | ||||||||||
| 2024-10-22 | OmniGen | BAAI | model | 4.33K | 17 | 1 | |||||||
| 2024-10-21 | Pangea | CMU | model | ||||||||||
| 2024-10-17 | Janus 4 | DeepSeek | model paper dataset | 17.75K | 10.89K | 11 | |||||||
| 2024-10-15 | LAPA | KAIST | model | ||||||||||
| 2024-10-15 | OKAMI | UT Austin | paper | ||||||||||
| 2024-10-15 | Zyda-2 | Zyphra | dataset | ||||||||||
| 2024-10-14 | Agent-as-a-Judge | KAUST, Meta | paper | ||||||||||
| 2024-10-14 | Deep Compression Autoencoder | MIT | model | ||||||||||
| 2024-10-14 | TULIP | University of Amsterdam | model | ||||||||||
| 2024-10-12 | OpenR | University College London | library | ||||||||||
| 2024-10-11 | Baichuan-Omni 2 | Baichuan | paper model | 273 | 7B | ||||||||
| 2024-10-10 | MathCoder2 / MathCode-Pile | CUHK | model | ||||||||||
| 2024-10-10 | SIFT (Active Fine-Tuning at Test Time) | ETH Zürich | paper | ||||||||||
| 2024-10-10 | Orthrus | Toronto & Vector Institute | model | ||||||||||
| 2024-10-10 | ★ RDT (Robotics Diffusion Transformer) | Tsinghua | model | 1.2B | |||||||||
| 2024-10-10 | GenARM | University of Maryland | paper | ||||||||||
| 2024-10-09 | REPA — Representation Alignment | NYU | paper | ||||||||||
| 2024-10-09 | MLE-bench | OpenAI | eval | 1.57K | 9 | 75 | |||||||
| 2024-10-09 | ★ PLaMo-100B | PFN | model | 118 | 100B | 2T | |||||||
| 2024-10-09 | ★ F5-TTS | Shanghai Jiao Tong University | model | ||||||||||
| 2024-10-09 | HELM | Stanford | eval | ||||||||||
| 2024-10-09 | Demis Hassabis & John Jumper Awarded Nobel Prize in Chemistry Google DeepMind | news | |||||||||||
| 2024-10-08 | LightRAG | HKU | library | ||||||||||
| 2024-10-07 | ★ Falcon Mamba | TII | model | 123.63K | 3 | 7.27B | 5.8T | 8.19K | |||||
| 2024-10-03 | LLaVA-Critic | NTU | model | ||||||||||
| 2024-10-03 | HELMET | Princeton | eval | ||||||||||
| 2024-10-03 | ProLong | Princeton | model | ||||||||||
| 2024-10-03 | LLMs Know More Than They Show | Technion | paper | ||||||||||
| 2024-10-03 | SageAttention | Tsinghua | library | ||||||||||
| 2024-10-02 | Knowledge Entropy Decay | KAIST | paper | ||||||||||
| 2024-10-02 | GFlowNets | Mila | paper | ||||||||||
| 2024-10-02 | VinePPO | Mila | paper | ||||||||||
| 2024-10-02 | ★ Llama-3.1-Nemotron-70B | NVIDIA | model | 58 | 4 | 128K | |||||||
| 2024-10-02 | Poolside Raises $500M Series B at ~$3B Valuation, Led by Bain Capital Ventures Crunchbase News | Poolside | news | ||||||||||
| 2024-10-01 | RATIONALYST | Johns Hopkins | paper | ||||||||||
| 2024-10-01 | TxT360 | MBZUAI | dataset | 22 | |||||||||
| 2024-10-01 | DSPy | Stanford | library | ||||||||||
| 2024-09-29 | Arc2Face | Imperial | model | ||||||||||
| 2024-09-28 | veRL (HybridFlow) | HKU | library | ||||||||||
| 2024-09-27 | ★ Emu3 2 | BAAI | paper | 2.42K | 4 | ||||||||
| 2024-09-27 | Realistic Evaluation of Model Merging | Toronto & Vector Institute | paper | ||||||||||
| 2024-09-25 | ★ Molmo | Ai2 | model | 913 | 2.17K | 8 | 6 (v4) | ||||||
| 2024-09-25 | ProX (Programming Every Example) | Shanghai Jiao Tong University | dataset | ||||||||||
| 2024-09-23 | ★ AMPLIFY | Mila | model | 350M | |||||||||
| 2024-09-23 | MobileUI Dataset | Xiaomi | dataset | 79 | |||||||||
| 2024-09-23 | MobileVLM | Xiaomi | model | 79 | |||||||||
| 2024-09-19 | ★ Qwen2.5 3 | Alibaba | model paper | 27.3K | 74 | 72B (max) | 18T | 8 (v4) | |||||
| 2024-09-19 | Scaling FP8 Training to Trillion-Token LLMs | Technion | paper | ||||||||||
| 2024-09-18 | Qwen2-VL | Alibaba | model | 27.42K | 76 | ||||||||
| 2024-09-18 | Qwen2.5-Coder 2 | Alibaba | model paper | 16.61K | 37 | 480B (max) | 7 (v4) | ||||||
| 2024-09-18 | Qwen2.5-Math 2 | Alibaba | model paper | 1.08K | 12 | 72B (max) | |||||||
| 2024-09-17 | Promptriever | Johns Hopkins | model | ||||||||||
| 2024-09-15 | Berkeley Function-Calling Leaderboard | UC Berkeley | eval | ||||||||||
| 2024-09-12 | ★ o1 | OpenAI | model | ~3.5T | 200K | 15 (v4) | |||||||
| 2024-09-11 | FuXi (weather & ocean FMs) | Fudan University | model | ||||||||||
| 2024-09-11 | ★ Pixtral 12B | Mistral | model | 4.18K | 6 | 12B | 128K | ||||||
| 2024-09-09 | Robot Utility Models | NYU | model | ||||||||||
| 2024-09-05 | ★ AdEMAMix Optimizer | Apple | paper | ||||||||||
| 2024-09-05 | ★ DeepSeek-V2.5 | DeepSeek | model | 7.71K | |||||||||
| 2024-09-05 | ★ MiniCPM3-4B | OpenBMB | model | 9.42K | 40.57K | 4B | |||||||
| 2024-09-05 | Open-MAGVIT2 | Tencent | library | 1.01K | |||||||||
| 2024-09-05 | FireRedTTS | Xiaohongshu | model | ||||||||||
| 2024-09-05 | Silvio Savarese Named to TIME 100 Most Influential in AI Salesforce | Salesforce | news | ||||||||||
| 2024-09-04 | MMMU-Pro | CMU | eval | 1.73K | |||||||||
| 2024-09-03 | ★ OLMoE | Ai2 | model | 1.03K | 103.29K | 7 | 6.9B | 1.3B | 5.13T | 4.1K | |||
| 2024-09-01 | MLC-LLM / WebLLM | CMU | library | ||||||||||
| 2024-08-31 | Hailuo AI (Video-01 / 2.3) | MiniMax | model | ||||||||||
| 2024-08-29 | CogVLM2 | Z.ai | model | 6 | |||||||||
| 2024-08-28 | Auxiliary-Loss-Free Load Balancing Strategy | DeepSeek | paper | 6 | |||||||||
| 2024-08-28 | STORM / Co-STORM | Stanford | library | ||||||||||
| 2024-08-26 | Fire-Flyer AI-HPC: Cost-Effective Software-Hardware Co-Design | DeepSeek | paper | ||||||||||
| 2024-08-22 | ★ Show-o | NUS | model | ||||||||||
| 2024-08-21 | Minitron | NVIDIA | paper | 380 | 3 | ||||||||
| 2024-08-21 | ★ Sarashina2 | SB Intuitions | model | 913 | 70B | 2.1T | 4.1K | ||||||
| 2024-08-15 | LiveCodeBench | UC Berkeley | eval | ||||||||||
| 2024-08-15 | Cybench | Stanford | eval | 40 | |||||||||
| 2024-08-13 | ★ SWE-bench (Verified / Multimodal / Multilingual) | Princeton | eval | 500 | |||||||||
| 2024-08-12 | MovieSum | University of Edinburgh | dataset | ||||||||||
| 2024-08-12 | ★ Tanuki | UTokyo | model | 47B | 13B | 1.7T | |||||||
| 2024-08-12 | CogVideoX: Text-to-Video Diffusion Models | Z.ai | model | 12.78K | 16 | ||||||||
| 2024-08-08 | Exascale Climate Emulator | KAUST | model | ||||||||||
| 2024-08-07 | ★ EXAONE 3.0 | LG | model | 41.67K | 2 | 7.8B | |||||||
| 2024-08-06 | LLaVA-OneVision | NTU | model | ||||||||||
| 2024-08-05 | MiniCPM-V 2.6 | OpenBMB | model | 25.58K | 23 | 8B | |||||||
| 2024-08-04 | scTab | TU Munich | paper | ||||||||||
| 2024-08-02 | GPUDrive | NYU | library | ||||||||||
| 2024-08-01 | Chatbot Arena | UC Berkeley, Stanford | eval | ||||||||||
| 2024-08-01 | EXAONEPath 1.0 | LG | paper | 1 | |||||||||
| 2024-08-01 | MiniMax Music Series | MiniMax | model | ||||||||||
| 2024-08-01 | KTransformers | Tsinghua | library | ||||||||||
| 2024-08-01 | Pinal / Denovo-Pinal | Westlake University | model | 16B | |||||||||
| 2024-07-31 | Large Language Monkeys | Stanford | paper | ||||||||||
| 2024-07-29 | ★ Apple Foundation Models (AFM) | Apple | paper | 4 | |||||||||
| 2024-07-29 | MindSearch | PJLab | library | 6.87K | 2 | ||||||||
| 2024-07-24 | ★ Mistral Large 2 | Mistral | model | 6.26K | 123B | 128K | 8 (v4) | ||||||
| 2024-07-24 | Model Collapse | University of Oxford | paper | ||||||||||
| 2024-07-23 | OpenHands | CMU | library | ||||||||||
| 2024-07-23 | ★ Meditron | EPFL | model | 70B (max) | |||||||||
| 2024-07-23 | AbdomenAtlas | Johns Hopkins | dataset | ||||||||||
| 2024-07-23 | ★ Llama 3.1 | Meta | model | 213.13K | 405B | 15.6T | 128K | 7 (v4) | |||||
| 2024-07-22 | RazorAttention: KV Cache Compression Through Retrieval Heads | Huawei | paper | ||||||||||
| 2024-07-21 | Dynamic Memory Compression | University of Edinburgh, NVIDIA | paper | ||||||||||
| 2024-07-21 | ★ Debating with More Persuasive LLMs Leads to More Truthful Answers | University College London | paper | ||||||||||
| 2024-07-21 | Clifford-Steerable CNNs | University of Amsterdam | paper | ||||||||||
| 2024-07-20 | Consent in Crisis: The Rapid Decline of the AI Data Commons | Cohere | paper | 11 | |||||||||
| 2024-07-20 | ★ Falcon 2 | TII | model | 4.48K | 2 | 11B | 5.5T | 8.19K | |||||
| 2024-07-18 | Mistral NeMo | Mistral | model | 32.02K | 12B | 128K | |||||||
| 2024-07-18 | SciCode | UIUC, CMU | eval | 338 | 80 | ||||||||
| 2024-07-17 | ★ LMMs-Eval | NTU | library | ||||||||||
| 2024-07-16 | Does Refusal Training Generalize to the Past Tense? | EPFL | paper | ||||||||||
| 2024-07-16 | BRIGHT | HKU | eval | 1.38K | |||||||||
| 2024-07-16 | Goldfish / MiniGPT4-video | KAUST | model | ||||||||||
| 2024-07-16 | Codestral Mamba | Mistral | model | 47.98K | 7.3B | ||||||||
| 2024-07-15 | sbi | Tübingen | library | ||||||||||
| 2024-07-11 | EchoMimicV2 & V3 | Ant Group | paper | 4.25K | 2 | ||||||||
| 2024-07-11 | FlashAttention-3 | Princeton | paper | ||||||||||
| 2024-07-11 | Skywork-Math | Skywork | paper | ||||||||||
| 2024-07-11 | Q-GaLore | UT Austin | library | ||||||||||
| 2024-07-10 | Training on the Test Task Confounds Evaluation | Tübingen | paper | ||||||||||
| 2024-07-08 | Anole | Shanghai Jiao Tong University | model | ||||||||||
| 2024-07-06 | Kolors 2 | Kuaishou | model paper | 4.61K | 401 | ||||||||
| 2024-07-05 | ★ SenseNova 5.5 | SenseTime | model | ||||||||||
| 2024-07-05 | Vimi | SenseTime | model | ||||||||||
| 2024-07-04 | MiniGPT-Med | KAUST | model | ||||||||||
| 2024-07-04 | ★ LLM-jp (v1/v2) | NII | model | 5 | 13B | ||||||||
| 2024-07-04 | ★ Step-2 | StepFun | model | 1T | |||||||||
| 2024-07-03 | LivePortrait 2 | Kuaishou | library paper | 18.53K | 7.37K | 9 | |||||||
| 2024-07-03 | ★ InternLM2.5 | PJLab | model | 7.22K | 106.78K | 1M | |||||||
| 2024-07-02 | InternVL 2.0 | PJLab | model | 10.06K | 305.47K | 108B (max) | |||||||
| 2024-07-01 | rsl_rl | ETH Zürich | library | ||||||||||
| 2024-07-01 | Copyright Traps for LLMs | Imperial | paper | ||||||||||
| 2024-07-01 | QPlanner | LG | paper | ||||||||||
| 2024-07-01 | Mathstral 7B | Mistral | model | 19.29K | 7B | ||||||||
| 2024-07-01 | MMLongBench-Doc | PJLab | dataset | 147 | 1 | ||||||||
| 2024-07-01 | ★ Agentless | UIUC | library | ||||||||||
| 2024-06-28 | ERNIE 4.0 Turbo | Baidu | model | ||||||||||
| 2024-06-28 | ★ YuLan | Renmin (RUC) | model | 12B | 1.7T | ||||||||
| 2024-06-26 | Zhang Hongjiang, founder of BAAI: 'AI systems should never be able to deceive humans' Financial Times | BAAI | news | ||||||||||
| 2024-06-25 | HEST-1k & TRIDENT | Harvard | dataset | ||||||||||
| 2024-06-24 | Mooncake 2 | Moonshot AI | paper dataset | 5.55K | 13 | ||||||||
| 2024-06-24 | ★ Cambrian-1 | NYU | model | ||||||||||
| 2024-06-24 | ★ Large Vocabulary Size Improves Large Language Models | SB Intuitions | paper | 1 | |||||||||
| 2024-06-21 | NAVSIM | Tübingen | eval | ||||||||||
| 2024-06-20 | ★ Claude 3.5 Sonnet | Anthropic | model | 200K | 8 (v4) | ||||||||
| 2024-06-19 | AgentDojo | ETH Zürich | eval | ||||||||||
| 2024-06-19 | ★ Semantic Entropy Hallucination Detection | University of Oxford | paper | ||||||||||
| 2024-06-18 | Transformer Neural Processes (TE-TNP / Gridded TNP) | University of Cambridge | paper | ||||||||||
| 2024-06-18 | OlympicArena | Shanghai Jiao Tong University | eval | 11.16K | |||||||||
| 2024-06-17 | AquilaMed-RL | BAAI | model | 24 | 1 | ||||||||
| 2024-06-17 | DeepSeek-Coder-V2 2 | DeepSeek | model paper | 6.83K | 4.01K | 48 | 6 (v4) | ||||||
| 2024-06-17 | Gaussian Splatting SLAM (MonoGS) | Imperial | library | ||||||||||
| 2024-06-17 | ★ Nemotron-4 340B | NVIDIA | model | 543 | 7 | 340B | 9T | 4.1K | |||||
| 2024-06-17 | Mip-Splatting | Tübingen | paper | ||||||||||
| 2024-06-17 | DataComp-LM (DCLM) | TU Munich | dataset | ||||||||||
| 2024-06-17 | GaussianAvatars | TU Munich | paper | ||||||||||
| 2024-06-17 | MeshGPT | TU Munich | paper | ||||||||||
| 2024-06-17 | ★ DataComp-LM (DCLM-7B) | University of Washington | model | 7B | |||||||||
| 2024-06-17 | MINT-1T | University of Washington | dataset | ||||||||||
| 2024-06-17 | 4K4D | Zhejiang University | model | ||||||||||
| 2024-06-14 | MASt3R | Naver | paper | 2.99K | 1 | ||||||||
| 2024-06-14 | Goldfish Loss | University of Maryland | paper | ||||||||||
| 2024-06-13 | 4M-21 | EPFL | model | ||||||||||
| 2024-06-13 | OpenVLA | Stanford, UC Berkeley | model | 7B | |||||||||
| 2024-06-12 | SciRIFF | Ai2 | dataset | 1 | |||||||||
| 2024-06-12 | Magpie | University of Washington | dataset | ||||||||||
| 2024-06-11 | Dasheng 3 | Xiaomi | paper model | 424 | 7B | ||||||||
| 2024-06-10 | LlamaGen | ByteDance | model | 1.95K | 5 | ||||||||
| 2024-06-10 | Language Models Resist Alignment | Peking University | paper | ||||||||||
| 2024-06-10 | PowerInfer | Shanghai Jiao Tong University | library | ||||||||||
| 2024-06-10 | Merlin | Stanford | model | ||||||||||
| 2024-06-10 | Apple Intelligence Introduced at WWDC 2024 Apple | Apple | news | ||||||||||
| 2024-06-09 | BiGGen Bench | KAIST | eval | 77 | |||||||||
| 2024-06-09 | Equivariant Neural Fields | University of Amsterdam | paper | ||||||||||
| 2024-06-06 | ★ Qwen2 2 | Alibaba | model paper | 27.3K | 49 | 72B (max) | 14B (max) | 6 (v4) | |||||
| 2024-06-06 | ★ Kling 2 | Kuaishou | model paper | ||||||||||
| 2024-06-06 | The Prompt Report | University of Maryland | paper | ||||||||||
| 2024-06-05 | GLM-4V | Z.ai | model | 2.33K | |||||||||
| 2024-06-04 | Seed-TTS 2 | ByteDance | model paper | 1.56K | 4 | ||||||||
| 2024-06-03 | ★ Skywork-MoE | Skywork | model | 140 | 556 | 3 | 146B | 22B | |||||
| 2024-06-01 | agentUniverse | Ant Group | library | 2.27K | 1 | ||||||||
| 2024-06-01 | FlagScale | BAAI | library | 518 | |||||||||
| 2024-06-01 | Yuan Embedding | Inspur | model | 3.53K | |||||||||
| 2024-06-01 | InfiniBench | KAUST | eval | ||||||||||
| 2024-06-01 | AtomGen & vector-inference 2 | Toronto & Vector Institute | model library | ||||||||||
| 2024-05-31 | ★ Mamba-2 | CMU, Princeton | paper | ||||||||||
| 2024-05-30 | MotionLLM 2 | IDEA Lab | model paper | 386 | 6 | ||||||||
| 2024-05-29 | ★ Codestral | Mistral | model | 13.02K | 22B | 32K | |||||||
| 2024-05-28 | Yuan 2.0-M32 | Inspur | model | 195 | 1.68K | 1 | 40B | 3.7B | |||||
| 2024-05-28 | Knowledge Circuits | Zhejiang University | paper | ||||||||||
| 2024-05-27 | RLAIF-V | OpenBMB | paper | 455 | 22 | 3 | |||||||
| 2024-05-24 | The Mosaic Memory of LLMs | Imperial | paper | ||||||||||
| 2024-05-23 | DeepSeek-Prover | DeepSeek | model | 577 | 118 | 6 | |||||||
| 2024-05-23 | SimPO | Princeton | paper | ||||||||||
| 2024-05-22 | ★ Baichuan 4 | Baichuan | model | ||||||||||
| 2024-05-22 | FlashRAG | Renmin (RUC) | library | ||||||||||
| 2024-05-21 | ★ Scaling Monosemanticity | Anthropic | paper | ||||||||||
| 2024-05-16 | ★ Grounding DINO 1.5 2 | IDEA Lab | model paper | 1.12K | 11 | ||||||||
| 2024-05-16 | OceanGPT | Zhejiang University | model | ||||||||||
| 2024-05-15 | ByteFF | ByteDance | model | 84 | |||||||||
| 2024-05-14 | Piccolo2 Embedding Model 2 | SenseTime | model paper | 145 | 43 | 1 | |||||||
| 2024-05-14 | Hunyuan-DiT 2 | Tencent | model paper | 4.29K | 1 | 1.5B | |||||||
| 2024-05-13 | AWQ | MIT | paper | ||||||||||
| 2024-05-13 | ★ GPT-4o | OpenAI | model | ~720B | 128K | 8 (v4) | |||||||
| 2024-05-13 | Plot2Code | Tencent | dataset | 24 | 1 | ||||||||
| 2024-05-13 | RLHFlow (RLHF Workflow) | UIUC | library | ||||||||||
| 2024-05-08 | AlphaFold 3 | paper | |||||||||||
| 2024-05-07 | ★ DeepSeek-V2 2 | DeepSeek | model paper | 5.01K | 5.35K | 102 | 236B | 21B | 6 (v4) | ||||
| 2024-05-07 | ★ Granite Code | IBM | model | 1.25K | 10 | 34B | 4.5T | ||||||
| 2024-05-07 | VeRA | University of Amsterdam | paper | ||||||||||
| 2024-05-07 | ★ SaProt | Westlake University | model | 1.3B (max) | |||||||||
| 2024-05-06 | SWE-agent | Princeton | library | ||||||||||
| 2024-05-02 | ★ Prometheus / Prometheus 2 | KAIST | model | ||||||||||
| 2024-05-01 | ReFT / pyreft | Stanford | library | ||||||||||
| 2024-05-01 | ProTrek | Westlake University | model | ||||||||||
| 2024-04-26 | llm-jp-corpus | NII | dataset | 47 | 4 | ||||||||
| 2024-04-25 | InternVL 1.5 | PJLab | model | 10.06K | 9.81K | 16 | 26B | ||||||
| 2024-04-25 | ShareGPT-4o | PJLab | dataset | 16 | |||||||||
| 2024-04-24 | ★ SenseNova 5.0 | SenseTime | model | 200K | |||||||||
| 2024-04-22 | OpenELM | Apple | model | 2 | 3B | ||||||||
| 2024-04-22 | SEED-X | Tencent | model | 558 | 1 | ||||||||
| 2024-04-18 | ★ Reka Core, Flash, and Edge | Reka | model | 125 | 3 | 67B | 128K | 6 (v4) | |||||
| 2024-04-17 | ★ ABAB 6 / 6.5 | MiniMax | model | ||||||||||
| 2024-04-17 | ★ Mixtral 8x22B | Mistral | model | 4.66K | 141B | 39B | 64K | ||||||
| 2024-04-16 | MiniCheck | UT Austin | library | ||||||||||
| 2024-04-12 | ★ MiniCPM-V 4 | OpenBMB | model paper | 25.58K | 163.09K | 23 | 9B | ||||||
| 2024-04-11 | ★ OSWorld | HKU | eval | 369 | |||||||||
| 2024-04-11 | MiniCPM-V 2.0 | OpenBMB | model | 67.98K | 2.8B | ||||||||
| 2024-04-11 | ★ Rho-1: Not All Tokens Are What You Need | PJLab, Microsoft | paper | ||||||||||
| 2024-04-03 | VAR (Visual Autoregressive Modeling) | ByteDance | model | 8.7K | 7 | ||||||||
| 2024-04-03 | PiSSA | Peking University | paper | ||||||||||
| 2024-04-02 | Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks | EPFL | paper | ||||||||||
| 2024-04-02 | ★ HyperCLOVA X | Naver | model | 6 | |||||||||
| 2024-04-01 | RULER: What's the Real Context Size of Your LLM? | NVIDIA | eval | 11 | 13 | ||||||||
| 2024-03-28 | ★ Jamba | AI21 Labs | model | 515 | 41 | 398B | 94B | 256K | 6 (v4) | ||||
| 2024-03-28 | Dataverse | Upstage | library | 564 | |||||||||
| 2024-03-28 | sDPO | Upstage | paper | 2 | |||||||||
| 2024-03-28 | Jamba: First Production-Grade SSM-Transformer Hybrid Released AI21 Labs | AI21 Labs | news | ||||||||||
| 2024-03-26 | 2D Gaussian Splatting | Tübingen | paper | ||||||||||
| 2024-03-25 | PairS | University of Cambridge | paper | ||||||||||
| 2024-03-23 | ★ Step-1 | StepFun | model | 130B | |||||||||
| 2024-03-23 | Step-1V / 1.5V / 2V | StepFun | model | ||||||||||
| 2024-03-23 | Understanding Emergent Abilities from the Loss Perspective | Z.ai | paper | 1 | |||||||||
| 2024-03-22 | UniTraj | EPFL | library | ||||||||||
| 2024-03-19 | MergeKit | Arcee | library | 7.13K | 3 | ||||||||
| 2024-03-19 | Dated Data | Johns Hopkins | paper | ||||||||||
| 2024-03-17 | ★ Grok-1 | SpaceXAI | model | 314B | 78B | 8.19K | 6 (v4) | ||||||
| 2024-03-12 | ★ Command R / R+ | Cohere | model | 33.71K | 104B | 128K | 5 (v4) | ||||||
| 2024-03-11 | ★ Stealing Part of a Production Language Model | ETH Zürich | paper | ||||||||||
| 2024-03-11 | Unraveling the Mystery of Scaling Laws: Part I | Meituan | paper | ||||||||||
| 2024-03-08 | DeepSeek-VL | DeepSeek | model | 4.13K | 8.76K | 45 | |||||||
| 2024-03-08 | CogView3 | Z.ai | model | 144 | |||||||||
| 2024-03-04 | ★ Claude 3 | Anthropic | model | 200K | 9 (v4) | ||||||||
| 2024-03-01 | Kimi 2M | Moonshot AI | model | 2M | |||||||||
| 2024-02-28 | WanJuan 2.0 (WanJuan-CC) | PJLab | dataset | ||||||||||
| 2024-02-27 | BioT5+ | Microsoft | paper | 4 | |||||||||
| 2024-02-26 | Craftax | University of Oxford | paper | ||||||||||
| 2024-02-26 | ★ scGPT 2 | Toronto & Vector Institute | model | ||||||||||
| 2024-02-23 | MegaScale | ByteDance | library | 24 | |||||||||
| 2024-02-22 | MATH-Vision (MATH-V) | CUHK | eval | 3.04K | |||||||||
| 2024-02-21 | SDXL-Lightning | ByteDance | model | 72.67K | 6 | ||||||||
| 2024-02-21 | ★ Gemma | model | 26.93K | 225 | 7B | ||||||||
| 2024-02-19 | AnyGPT | Fudan University | model | ||||||||||
| 2024-02-15 | ★ Gemini 1.5 Pro | model | 282 | 1M | 8 (v4) | ||||||||
| 2024-02-15 | SAMformer 2 | Huawei | model paper | 190 | 1 | ||||||||
| 2024-02-15 | ★ Sora | OpenAI | model | ||||||||||
| 2024-02-14 | Soft Prompt Threats | TU Munich | paper | ||||||||||
| 2024-02-07 | ★ Moirai | Salesforce | model | 1.52K | 109.96K | 32 | 311M | ||||||
| 2024-02-06 | ★ SenseNova 4.0 | SenseTime | model | ||||||||||
| 2024-02-05 | ★ BGE-M3 2 | BAAI | model paper | 11.8K | 28.87M | 52 | |||||||
| 2024-02-05 | DeepSeek-Math 2 | DeepSeek | model paper | 3.32K | 69 | ||||||||
| 2024-02-04 | ★ Qwen1.5 | Alibaba | model | 110B (max) | 6 (v4) | ||||||||
| 2024-02-04 | Aligner | Peking University | paper | ||||||||||
| 2024-02-01 | ★ OLMo | Ai2 | model | 6.53K | 2.4K | 9 | 7B | 2.46T | 2.05K | ||||
| 2024-02-01 | ★ Aya 101 | Cohere | model | 8.81K | 12 | 13B | |||||||
| 2024-02-01 | ★ MiniCPM 3 | OpenBMB | model paper | 9.42K | 3.73K | 20 | 2B (max) | ||||||
| 2024-01-31 | ★ Dolma | Ai2 | dataset | 1.51K | 9 | ||||||||
| 2024-01-30 | YOLO-World | Tencent | paper | 6.4K | 28 | ||||||||
| 2024-01-30 | infini-gram | University of Washington | library | ||||||||||
| 2024-01-29 | ★ Baichuan 3 | Baichuan | model | ||||||||||
| 2024-01-23 | ★ InternLM2 2 | PJLab | model paper | 7.22K | 19.21K | 27 | |||||||
| 2024-01-22 | Binoculars | University of Maryland | library | ||||||||||
| 2024-01-20 | TFLOP | Upstage | paper | 51 | |||||||||
| 2024-01-19 | Depth Anything | ByteDance | model | 8.26K | 22 | ||||||||
| 2024-01-17 | ★ AlphaGeometry | paper | |||||||||||
| 2024-01-17 | ★ GLM-4 | Z.ai | model | 7.07K | |||||||||
| 2024-01-15 | SciGLM / SciInstruct | Z.ai | paper | 4 | |||||||||
| 2024-01-11 | ★ DeepSeek-MoE 2 | DeepSeek | model paper | 18.65K | 16 | 16B | 2.8B | ||||||
| 2024-01-09 | Lightning Linear Attention | Ant Group | paper | 2 | |||||||||
| 2024-01-09 | Baichuan-NPC | Baichuan | model | ||||||||||
| 2024-01-04 | LLaMA Pro | Tencent | model | 513 | 1.37K | 1 | |||||||
| 2024-01-02 | ★ EasyEdit | Zhejiang University | library | ||||||||||
| 2024-01-01 | VSAG | Ant Group | library | 482 | |||||||||
| 2024-01-01 | FlagAI | BAAI | library | 3.87K | |||||||||
| 2024-01-01 | BenchMARL | University of Cambridge, Meta | library | ||||||||||
| 2024-01-01 | scikit-learn | INRIA | library | ||||||||||
| 2023-12-28 | PanGu-pi 3 | Huawei | model paper | 3.16K | 2 | 7B (max) | |||||||
| 2023-12-28 | ★ Spike No More: Stabilizing the Pre-training of Large Language Models | SB Intuitions | paper | 2 | |||||||||
| 2023-12-23 | ★ SOLAR 10.7B | Upstage | model | 51.95K | 7 | 10.7B | 6 (v4) | ||||||
| 2023-12-22 | GraphCast | paper | 6.67K | 170 | |||||||||
| 2023-12-21 | DUSt3R | Naver | paper | 7.19K | 3 | ||||||||
| 2023-12-21 | ★ InternVL: Scaling up Vision Foundation Models | PJLab | model | 10.06K | 17 | 6B | |||||||
| 2023-12-20 | ★ Emu2 | BAAI | model | 1.77K | 564 | 7 | |||||||
| 2023-12-16 | Paloma | Ai2 | dataset | ||||||||||
| 2023-12-11 | ★ Mixtral 8x7B | Mistral | model | 60.57K | 120 | 46.7B | 12.9B | 32K | 5 (v4) | ||||
| 2023-12-06 | ★ Gemini 1.0 | model | 809 | 32K | 6 (v4) | ||||||||
| 2023-12-05 | ★ MLX | Apple | library | 26.82K | |||||||||
| 2023-12-05 | Lenna 2 | Meituan | model paper | 87 | 1 | ||||||||
| 2023-12-05 | ReasonDet | Meituan | dataset | 87 | 1 | ||||||||
| 2023-12-04 | Magicoder / OSS-Instruct | UIUC | model | ||||||||||
| 2023-11-29 | ★ DeepSeek-LLM 2 | DeepSeek | model paper | 7.03K | 1.57K | 87 | 67B | 5 (v4) | |||||
| 2023-11-29 | GNoME (Materials Discovery) | paper | 1.19K | ||||||||||
| 2023-11-28 | ★ Falcon (7B / 40B / 180B) | TII | model | 31 | 115 | 180B | 3.5T | 2.05K | |||||
| 2023-11-27 | MagicAnimate & Make Pixels Dance | ByteDance | model | 10.91K | 7 | ||||||||
| 2023-11-27 | Yuan 2.0 | Inspur | model | 688 | 102.6B (max) | ||||||||
| 2023-11-27 | UniRepLKNet | Tencent | model | 1.07K | 34 | ||||||||
| 2023-11-22 | T-Rex 3 | IDEA Lab | model paper | 2.68K | 8 | ||||||||
| 2023-11-20 | ★ GPQA: Graduate-Level Google-Proof Q&A | Anthropic, Ai2 | eval | 510 | 21 | 198 | |||||||
| 2023-11-16 | JaxMARL | University of Oxford | library | ||||||||||
| 2023-11-14 | Qwen2-Audio 2 | Alibaba | model paper | 2.08K | 1.48K | 27 | 7B (max) | ||||||
| 2023-11-06 | CogVLM | Z.ai | model | 6.74K | 77 | ||||||||
| 2023-11-02 | DeepSeek-Coder 2 | DeepSeek | model paper | 23.66K | 6.35K | 107 | 33B (max) | ||||||
| 2023-10-30 | ★ SEA-LION v1 | AI Singapore | model | 7B | 980B | ||||||||
| 2023-10-30 | ★ Skywork-13B | Skywork | model | 729 | 11 | 13B | 3.2T | 4.1K | |||||
| 2023-10-25 | DiQAD | Baidu | dataset | 1 | |||||||||
| 2023-10-19 | KwaiYiiMath 2 | Kuaishou | model paper | ||||||||||
| 2023-10-17 | ★ ERNIE 4.0 | Baidu | model | ||||||||||
| 2023-10-17 | ★ BitNet | Microsoft | paper | 63 | |||||||||
| 2023-10-13 | VideoCrafter | Tencent | library | 5.06K | |||||||||
| 2023-10-12 | ★ Aquila2 | BAAI | model | 445 | 36 | 34B (max) | |||||||
| 2023-10-09 | ★ Kimi-v1 | Moonshot AI | model | 200K | |||||||||
| 2023-10-05 | ★ MathCoder | PJLab | paper | 339 | 30 | 4 | |||||||
| 2023-10-04 | SEED / SEED-LLaMA | Tencent | model | 641 | 18 | ||||||||
| 2023-10-03 | LanguageBind / Video-LLaVA | Peking University | model | ||||||||||
| 2023-10-01 | DALL-E 3 | OpenAI | model | ||||||||||
| 2023-09-29 | ToRA: Tool-Integrated Reasoning Agent | Microsoft | paper | 1.12K | 21 | ||||||||
| 2023-09-28 | PLaMo-13B | PFN | model | 162 | 13B | 4.1K | |||||||
| 2023-09-27 | ★ Mistral 7B | Mistral | model | 10.81K | 479.84K | 288 | 7.3B | 32K | 5 (v4) | ||||
| 2023-09-26 | InternLM-XComposer | PJLab | model | 2.92K | 13.08K | 31 | |||||||
| 2023-09-25 | Qwen-Agent | Alibaba | library | 16.51K | |||||||||
| 2023-09-25 | qwen.cpp | Alibaba | library | 627 | |||||||||
| 2023-09-21 | ★ PengCheng-Mind | PCL | model | 6 | 4 | 200B | 1.5T | ||||||
| 2023-09-12 | vLLM (original release) | UC Berkeley | paper | ||||||||||
| 2023-09-11 | NExT-GPT | NUS | model | ||||||||||
| 2023-09-08 | Ant Financial LLM | Ant Group | model | ||||||||||
| 2023-09-08 | CodeFuse | Ant Group | model | ||||||||||
| 2023-09-08 | Fin-Eval | Ant Group | dataset | ||||||||||
| 2023-09-07 | ★ Hunyuan-LLM | Tencent | model | ||||||||||
| 2023-09-06 | ★ Baichuan 2 3 | Baichuan | paper model | 4.1K | 210.77K | 125 | 13B | 2.6T | |||||
| 2023-09-01 | OpenSPG & OpenAGL | Ant Group | library | 2.12K | |||||||||
| 2023-09-01 | Asclepius | KAIST | model | ||||||||||
| 2023-09-01 | XTuner | PJLab | library | 5.15K | |||||||||
| 2023-08-31 | ★ ERNIE 3.5 | Baidu | model | ||||||||||
| 2023-08-31 | Belebele | Meta | eval | 1 | 109.8K | 122 | |||||||
| 2023-08-30 | ★ JAIS | MBZUAI | model | 1.06K | 23 | 30B | 1.63T | ||||||
| 2023-08-29 | DISC-MedLLM | Fudan University | model | 13B | |||||||||
| 2023-08-29 | LongBench | Z.ai | eval | 1.19K | 10 | 4.75K | 21 | ||||||
| 2023-08-24 | Qwen-VL | Alibaba | model | 136.35K | 139 | ||||||||
| 2023-08-24 | Code Llama | Meta | model | 398 | 70B | ||||||||
| 2023-08-22 | Lagent & AgentLego | PJLab | library | 2.26K | |||||||||
| 2023-08-21 | WanJuan 1.0 Corpus | PJLab | dataset | 8 | |||||||||
| 2023-08-20 | ViT-Lens | Tencent | paper | 190 | 1 | ||||||||
| 2023-08-18 | KwaiYii | Kuaishou | model | 175B | |||||||||
| 2023-08-15 | ★ Aquila 2 | BAAI | model paper | 445 | 1.06K | 33B (max) | |||||||
| 2023-08-11 | MiLM-6B | Xiaomi | model | 458 | 6B | ||||||||
| 2023-08-08 | Baichuan-53B | Baichuan | model | 53B | |||||||||
| 2023-08-04 | Weblab-10B | UTokyo | model | 10B | 600B | ||||||||
| 2023-08-04 | SoftBank launches an OpenAI for Japan: SB Intuitions, building LLMs and generative AI in Japanese TechCrunch | SB Intuitions | news | ||||||||||
| 2023-08-03 | ★ Qwen 3 | Alibaba | model paper | 21.27K | 18.12K | 89 | 72B (max) | 5 (v4) | |||||
| 2023-08-02 | ★ BGE Text Embeddings | BAAI | model | 11.8K | 13.9M | ||||||||
| 2023-08-01 | FlagEmbedding & C-MTEB 2 | BAAI | library dataset | 11.8K | 70 | ||||||||
| 2023-08-01 | ★ ABAB 5 / 5.5 | MiniMax | model | ||||||||||
| 2023-07-31 | ToolLLM: Facilitating LLMs to Master 16000+ APIs | OpenBMB | paper | 5.66K | 69 | ||||||||
| 2023-07-30 | SEED-Bench | Tencent | eval | 364 | 58 | 19.24K | 12 | ||||||
| 2023-07-27 | GCG (Universal Adversarial Attacks) | CMU | paper | ||||||||||
| 2023-07-20 | FLASK | KAIST | paper | ||||||||||
| 2023-07-19 | ★ EXAONE 2.0 | LG | model | ||||||||||
| 2023-07-18 | ★ Llama 2 | Meta | model | 11 | 2.62K | 70B | 2T | 4.1K | 5 (v4) | ||||
| 2023-07-16 | ChatDev 2 | OpenBMB | paper library | 33.36K | 69 | ||||||||
| 2023-07-13 | InternVid | PJLab | dataset | 2.28K | 32 | ||||||||
| 2023-07-11 | ★ Emu | BAAI | model | 1.77K | 29 | ||||||||
| 2023-07-11 | Baichuan-13B | Baichuan | model | 2.93K | 14.72K | 13B | |||||||
| 2023-07-08 | Ant Group Proposes Buying Back up to 7.6% of Its Shares at a ¥567.1B (~$78.5B) Valuation, ~70% Below Its 2020 IPO Mark Fortune (Bloomberg) | Ant Group | news | ||||||||||
| 2023-07-07 | ★ SenseNova 2.0 Upgrade | SenseTime | model | ||||||||||
| 2023-07-06 | ★ InternLM-1.0 | PJLab | model | 7.22K | 1.68K | 104B | |||||||
| 2023-07-05 | PanGu-Weather 2 | Huawei | model paper | 1.36K | 125 | ||||||||
| 2023-07-01 | InternEvo | PJLab | library | 420 | |||||||||
| 2023-07-01 | OpenCompass | PJLab | library | 7.08K | |||||||||
| 2023-06-25 | ChatGLM2 / ChatGLM3 | Z.ai | model | 13.68K | 124.79K | 6B | |||||||
| 2023-06-23 | MME (Multimodal Evaluation) | BAAI | eval | 17.87K | 2.37K | 14 | |||||||
| 2023-06-21 | LMFlow | HKUST | library | ||||||||||
| 2023-06-20 | ★ Phi-1 ("Textbooks Are All You Need") | Microsoft | model | 99 | 1.3B | ||||||||
| 2023-06-20 | UniAD | SenseTime, PJLab | paper | 4.64K | 1 | ||||||||
| 2023-06-17 | LMQL | ETH Zürich | library | ||||||||||
| 2023-06-15 | ★ Baichuan-7B | Baichuan | model | 5.65K | 148.39K | 7B | 1.2T | ||||||
| 2023-06-14 | WebGLM | Z.ai | paper | 1.6K | 2 | ||||||||
| 2023-06-12 | detrex 2 | IDEA Lab | library paper | 2.29K | 15 | ||||||||
| 2023-06-07 | AlphaDev | paper | |||||||||||
| 2023-06-05 | LIBERO | UT Austin | eval | ||||||||||
| 2023-06-01 | DB-GPT | Ant Group | library | 18.97K | |||||||||
| 2023-06-01 | DLRover | Ant Group | library | 1.66K | |||||||||
| 2023-06-01 | LMDeploy | PJLab | library | 7.9K | |||||||||
| 2023-06-01 | ★ RefinedWeb | TII | dataset | 156 | |||||||||
| 2023-05-31 | Let's Verify Step by Step | OpenAI | paper | 2.14K | 30 | ||||||||
| 2023-05-30 | GPT4Tools | Tencent | paper | 770 | 33 | ||||||||
| 2023-05-29 | Mix-of-Show | Tencent | paper | 431 | 30 | ||||||||
| 2023-05-27 | CPM-Bee | OpenBMB | model | 2.41K | 3 | 10B | |||||||
| 2023-05-22 | ★ Grouped Query Attention (GQA) | paper | 27 | ||||||||||
| 2023-05-20 | PengCheng-Nebula | PCL | announcement | ||||||||||
| 2023-05-17 | ★ Ziya LLM 4 | IDEA Lab | model | 4.13K | 1.83K | ||||||||
| 2023-05-15 | C-Eval | HKUST | eval | 13.95K | |||||||||
| 2023-05-10 | PaLM 2 | model | ~340B | 5 (v4) | |||||||||
| 2023-05-05 | Otter (LMM) | NTU | model | ||||||||||
| 2023-05-04 | ★ StarCoder | ServiceNow | model | 22.54K | 192 | 15.5B | 1T | 8.19K | |||||
| 2023-05-02 | EvalPlus | UIUC | eval | ||||||||||
| 2023-04-20 | ★ MiniGPT-4 | KAUST | model | ||||||||||
| 2023-04-20 | UltraChat & UltraFeedback | OpenBMB | dataset | 2.86K | |||||||||
| 2023-04-12 | Edit-Friendly DDPM | Technion | paper | ||||||||||
| 2023-04-11 | SenseChat / SenseNova Launch | SenseTime | model | ||||||||||
| 2023-04-11 | SenseMirage | SenseTime | model | 10B (max) | |||||||||
| 2023-04-10 | Stable-DINO 2 | IDEA Lab | library paper | 242 | 5 | ||||||||
| 2023-04-06 | Grounded SAM 3 | IDEA Lab | library paper | 17.63K | 90 | ||||||||
| 2023-04-05 | ★ Segment Anything (SAM) | Meta | model | 54.33K | 538 | ||||||||
| 2023-04-01 | BMTools | OpenBMB | library | 2.77K | |||||||||
| 2023-03-31 | A Survey of Large Language Models | Renmin (RUC) | paper | ||||||||||
| 2023-03-27 | EVA-CLIP | BAAI | model | 2.68K | 80 | ||||||||
| 2023-03-27 | Qianfan Platform | Baidu | announcement | ||||||||||
| 2023-03-20 | ★ PanGu-Sigma 2 | Huawei | model paper | 7 | 1.1T | 329B | |||||||
| 2023-03-14 | OpenSeeD 2 | IDEA Lab | library paper | 759 | 2 | ||||||||
| 2023-03-14 | ★ GPT-4 | OpenAI | model | ~666B | 128K | 7 (v4) | |||||||
| 2023-03-14 | ★ ChatGLM-6B | Z.ai | model | 41.05K | 1.28K | 176 | 6B | ||||||
| 2023-03-09 | ★ Grounding DINO 2 | IDEA Lab | model paper | 10.25K | 2.22M | 245 | |||||||
| 2023-02-27 | ★ LLaMA | Meta | model | 3.9K | 65B | ||||||||
| 2023-02-20 | MOSS | Fudan University | model | 16B | |||||||||
| 2023-01-30 | ★ BLIP-2 | Salesforce | model | 581.31K | 914 | ||||||||
| 2023-01-23 | Microsoft Extends Multibillion-Dollar OpenAI Partnership Microsoft | Microsoft | news | ||||||||||
| 2023-01-01 | FlagEvaluation | BAAI | library | 13 | |||||||||
| 2022-12-22 | Tune-A-Video | Tencent | paper | 4.37K | 27 | ||||||||
| 2022-12-19 | Instructor | HKU | model | ||||||||||
| 2022-12-15 | ★ Constitutional AI | Anthropic | paper | 306 | |||||||||
| 2022-12-06 | InternVideo / InternVideo2 | PJLab | model | 2.28K | 93 | ||||||||
| 2022-12-05 | Painter | BAAI | model | 2.6K | 10 | ||||||||
| 2022-11-30 | ★ Speculative Decoding | paper | 34 | ||||||||||
| 2022-11-12 | AltCLIP & AltDiffusion | BAAI | model | 3.87K | 111.27K | 10 | |||||||
| 2022-11-10 | InternImage 2 | PJLab | model paper | 2.83K | 41 | ||||||||
| 2022-11-02 | Chinese CLIP | Alibaba | model | 5.93K | 53 | ||||||||
| 2022-11-02 | Taiyi 3 | IDEA Lab | model paper | 419 | 2 | ||||||||
| 2022-10-06 | ByteTransformer | ByteDance | library | 479 | 1 | ||||||||
| 2022-09-30 | CodeGeeX 2 | Z.ai | model paper | 8.79K | 49 | ||||||||
| 2022-09-21 | ★ Whisper | OpenAI | model | 102.39K | 5.05M | 1.16K | 1.55B | ||||||
| 2022-09-16 | CPM-Ant | OpenBMB | model | 500 | 10B | ||||||||
| 2022-09-03 | TuGraph | Ant Group | library | 1.74K | |||||||||
| 2022-08-24 | ★ GLM-130B 2 | Z.ai | model paper | 7.65K | 296 | 130B | |||||||
| 2022-08-01 | COYO-700M | Kakao | dataset | 1.26K | |||||||||
| 2022-07-22 | PanGu-Coder 3 | Huawei | model paper | 3.16K | 5 | 36 | 2.6B (max) | ||||||
| 2022-07-11 | Orca (continuous batching) | SNU | paper | ||||||||||
| 2022-07-04 | SecretFlow | Ant Group | library | 2.67K | |||||||||
| 2022-06-24 | YOLOv6 2 | Meituan | library paper | 5.89K | 1.75K | ||||||||
| 2022-06-06 | Mask DINO 2 | IDEA Lab | library paper | 1.54K | 20 | ||||||||
| 2022-06-01 | Vision GNN (ViG) 2 | Huawei | model paper | 4.42K | 194 | ||||||||
| 2022-05-26 | Matryoshka Representation Learning | University of Washington | paper | ||||||||||
| 2022-05-02 | ★ OPT (Open Pre-trained Transformer) | Meta | model | 175B | |||||||||
| 2022-04-26 | CogView2 | Z.ai, BAAI | model | 955 | 121 | ||||||||
| 2022-04-12 | ★ Training a Helpful and Harmless Assistant (HH-RLHF) | Anthropic | paper | 365 | |||||||||
| 2022-04-05 | ★ PaLM | model | 2.13K | 540B | |||||||||
| 2022-03-29 | ★ Chinchilla (Compute-Optimal Training) | paper | 663 | ||||||||||
| 2022-03-25 | ★ CodeGen | Salesforce | model | 5.17K | 1.81K | 235 | 16.1B | ||||||
| 2022-03-20 | Delta Tuning 2 | OpenBMB | paper library | 1.04K | |||||||||
| 2022-03-07 | DINO (DETR) 2 | IDEA Lab | library paper | 2.81K | 760 | ||||||||
| 2022-03-04 | ★ InstructGPT (RLHF) | OpenAI | paper | 4.29K | |||||||||
| 2022-03-02 | DN-DETR 2 | IDEA Lab | library paper | 605 | 55 | ||||||||
| 2022-02-11 | BMTrain | OpenBMB | library | 624 | |||||||||
| 2022-02-08 | ★ AlphaCode | paper | |||||||||||
| 2022-02-07 | OFA: One For All | Alibaba | model | 2.56K | 258 | ||||||||
| 2022-01-30 | FEDformer 2 | Huawei | model paper | 802 | 541 | ||||||||
| 2022-01-28 | ★ Chain-of-Thought Prompting | paper | 4.25K | ||||||||||
| 2022-01-28 | DAB-DETR 2 | IDEA Lab | library paper | 579 | 397 | ||||||||
| 2022-01-25 | SPIRAL 2 | Huawei | model paper | 604 | 7 | ||||||||
| 2022-01-24 | SenseCore AI Infrastructure | SenseTime | announcement | ||||||||||
| 2022-01-14 | DeepSpeed-MoE | Microsoft | paper | 42.49K | 55 | ||||||||
| 2021-12-31 | ERNIE-ViLG | Baidu | model | 30 | 10B | ||||||||
| 2021-12-01 | ★ EXAONE 1.0 | LG | model | 300B | |||||||||
| 2021-12-01 | tFold | Tencent | model | 158 | |||||||||
| 2021-11-30 | Donut | Naver | paper | 6.88K | 5 | ||||||||
| 2021-11-22 | Fengshenbang 3 | IDEA Lab | library model | 4.13K | 3.14K | ||||||||
| 2021-11-01 | KoGPT | Kakao | model | 1.01K | 17 | 6.2B | 2.05K | ||||||
| 2021-10-13 | ByteTrack | ByteDance | library | 6.44K | 105 | ||||||||
| 2021-10-10 | Yuan 1.0 | Inspur | model | 589 | 25 | 245B | |||||||
| 2021-09-28 | DiffVC 2 | Huawei | model paper | 604 | 25 | ||||||||
| 2021-09-10 | ★ HyperCLOVA | Naver | model | 4 | 204B | 560B | |||||||
| 2021-09-03 | ★ FLAN (Instruction Tuning) | paper | 69 | ||||||||||
| 2021-08-10 | ★ Codex | OpenAI | model | 3.26K | 1.43K | 12B | |||||||
| 2021-07-28 | Triton | OpenAI | library | 19.41K | |||||||||
| 2021-07-15 | ★ AlphaFold 2 | paper | |||||||||||
| 2021-07-12 | SPLADE | Naver | paper | 2 | |||||||||
| 2021-07-08 | OpenDILab / DI-engine 2 | SenseTime, PJLab | library | 3.62K | |||||||||
| 2021-07-08 | OpenPPL (PPLNN) | SenseTime | library | 1.37K | |||||||||
| 2021-07-07 | ★ HumanEval | OpenAI | eval | 3.26K | 1.43K | 164 | |||||||
| 2021-07-05 | ★ ERNIE 3.0 & 3.0 Titan | Baidu | model | 195 | 260B | ||||||||
| 2021-07-01 | Meituan Sky Project | Meituan | announcement | ||||||||||
| 2021-06-28 | PFP / Matlantis | PFN | model | ||||||||||
| 2021-06-24 | CPM-2 | BAAI | model | 164 | 15 | 198B | |||||||
| 2021-06-17 | ★ LoRA (Low-Rank Adaptation) | Microsoft | paper | 2.45K | |||||||||
| 2021-06-01 | OceanBase | Ant Group | library | 10.15K | |||||||||
| 2021-06-01 | ★ Wu Dao 2.0 2 | BAAI | model paper | 1.75T | |||||||||
| 2021-06-01 | Wu Dao Corpora | BAAI | dataset | ||||||||||
| 2021-05-26 | CogView | Z.ai, BAAI | model | 1.8K | 383 | ||||||||
| 2021-05-13 | Grad-TTS 2 | Huawei | model paper | 604 | 43 | ||||||||
| 2021-05-01 | Trustworthy AI White Paper | Xiaomi | paper | ||||||||||
| 2021-04-29 | ★ DINO | Meta | paper | ||||||||||
| 2021-04-26 | ★ PanGu-alpha 2 | Huawei, PCL | model paper | 3.16K | 94 | 200B | |||||||
| 2021-04-01 | SapBERT | University of Cambridge | model | ||||||||||
| 2021-03-20 | ★ Wu Dao 1.0 | BAAI | model | 11.3B (max) | |||||||||
| 2021-03-18 | ★ GLM (Original) 2 | Z.ai | model paper | 3.51K | 285 | 21 | 10B | ||||||
| 2021-03-18 | P-Tuning | Z.ai | paper | 2.08K | 265 | ||||||||
| 2021-03-01 | M6 Series | Alibaba | model | 48 | |||||||||
| 2021-02-27 | Transformer in Transformer (TNT) 2 | Huawei | model paper | 4.42K | 1.01K | ||||||||
| 2021-02-26 | ★ CLIP | OpenAI | paper | 33.74K | 5.3K | ||||||||
| 2021-01-11 | ★ Switch Transformer | paper | 361 | ||||||||||
| 2021-01-05 | ★ DALL-E | OpenAI | model | 1.13K | 12B | ||||||||
| 2021-01-01 | MS-MARCO-CN | Baidu | dataset | 18 | |||||||||
| 2021-01-01 | PaddleNLP | Baidu | library | 12.95K | |||||||||
| 2021-01-01 | PaddleSpeech | Baidu | library | 12.61K | |||||||||
| 2021-01-01 | KoBART | SK Telecom | model | 467 | 127 | ||||||||
| 2020-12-07 | HEBO 2 | Huawei | library paper | 2.77K | 19 | ||||||||
| 2020-12-01 | CPM-1 | BAAI | model | 1.58K | 22 | 2.6B | |||||||
| 2020-10-22 | ★ Vision Transformer (ViT) | paper | 12.58K | 21.58K | |||||||||
| 2020-10-08 | Deformable DETR 2 | SenseTime | model paper | 3.98K | 12.41K | 1.87K | |||||||
| 2020-09-12 | FuxiCTR / BARS 3 | Huawei | library paper | 1.43K | 12 | ||||||||
| 2020-08-01 | Vega | Huawei | library | 848 | |||||||||
| 2020-07-15 | PaddleOCR | Baidu | library | 81.73K | |||||||||
| 2020-06-11 | ★ GPT-3 | OpenAI | model | 3.03K | 175B | 300B | 2.05K | ||||||
| 2020-06-08 | ★ Liquid Time-constant Networks | Liquid AI | paper | 3 | |||||||||
| 2020-04-08 | DynaBERT 2 | Huawei | model paper | 3.16K | 12 | 119 | |||||||
| 2020-03-28 | MindSpore | Huawei | library | 4.69K | |||||||||
| 2020-03-13 | ★ ProGen | Salesforce | model | 702 | 34 | 1.2B | |||||||
| 2020-03-10 | Bolt | Huawei | library | 958 | |||||||||
| 2020-02-13 | ★ DeepSpeed | Microsoft | library | 42.49K | |||||||||
| 2020-01-23 | ★ Scaling Laws for Neural Language Models | OpenAI | paper | 1.51K | |||||||||
| 2020-01-01 | Kunlun XPU | Baidu | announcement | ||||||||||
| 2020-01-01 | PaddleDetection | Baidu | library | 14.24K | |||||||||
| 2020-01-01 | PaddleSeg | Baidu | library | 9.34K | |||||||||
| 2020-01-01 | KoGPT2 | SK Telecom | model | 558 | |||||||||
| 2019-11-27 | GhostNet 4 | Huawei | model paper | 4.42K | 406 | ||||||||
| 2019-09-23 | TinyBERT 2 | Huawei | model paper | 3.16K | 169.89K | 137 | |||||||
| 2019-09-17 | ★ Megatron-LM | NVIDIA | library | 16.66K | 835 | ||||||||
| 2019-09-01 | NEZHA 2 | Huawei | model paper | 3.16K | 86 | ||||||||
| 2019-08-23 | Ascend 910 Series | Huawei | announcement | ||||||||||
| 2019-08-01 | Chainer | PFN | library | 27 | |||||||||
| 2019-07-29 | ★ ERNIE 2.0 | Baidu | model | 74 | |||||||||
| 2019-07-25 | ★ Optuna | PFN | library | 14.34K | |||||||||
| 2019-07-19 | SUMBT | SK Telecom | paper | 90 | 24 | ||||||||
| 2019-06-20 | Alchemy | Tencent | dataset | 114 | 65 | ||||||||
| 2019-06-01 | KoBERT | SK Telecom | model | 1.41K | 18.01K | ||||||||
| 2019-04-19 | ★ ERNIE 1.0 2 | Baidu | model paper | 770 | |||||||||
| 2019-04-03 | CRAFT | Naver | paper | 3.38K | 58 | ||||||||
| 2019-02-14 | ★ GPT-2 | OpenAI | model | 24.92K | 13.36M | 1.5B | 1.02K | ||||||
| 2018-12-01 | ★ JAX | library | 35.79K | ||||||||||
| 2018-10-11 | ★ BERT | model | 340M | ||||||||||
| 2018-10-01 | OpenMMLab / MMDetection 2 | SenseTime, PJLab | library paper | 32.75K | 794 | ||||||||
| 2018-06-11 | ★ GPT-1 | OpenAI | model | 117M | 512 | ||||||||
| 2018-06-01 | MACE (Mobile AI Compute Engine) | Xiaomi | library | 5.04K | |||||||||
| 2018-03-16 | ApolloScape | Baidu | dataset | 617 | |||||||||
| 2018-03-01 | Kata Containers | Ant Group | library | 8.05K | |||||||||
| 2017-11-24 | StarGAN | Naver | paper | 18 | |||||||||
| 2017-11-14 | DuReader | Baidu | dataset | 1.18K | 51 | ||||||||
| 2017-11-02 | ★ VQ-VAE | paper | 1.93K | ||||||||||
| 2017-07-26 | Xiao AI | Xiaomi | announcement | ||||||||||
| 2017-07-20 | ★ Proximal Policy Optimization (PPO) | OpenAI | paper | ||||||||||
| 2017-07-05 | Apollo | Baidu | library | 26.66K | |||||||||
| 2017-06-12 | ★ Attention Is All You Need (Transformer) | paper | |||||||||||
| 2017-06-01 | Angel ML | Tencent | library | 6.79K | |||||||||
| 2017-05-03 | AI.SG: New National Programme to Catalyse, Synergise and Boost Singapore's AI Capabilities (NRF to Invest Up to S$150M Over Five Years) Ministry of Digital Development and Information | AI Singapore | news | ||||||||||
| 2017-03-15 | DiscoGAN | SK Telecom | paper | 776 | 732 | ||||||||
| 2017-01-18 | ★ PyTorch | Meta | library | 100.65K | 16.19K | ||||||||
| 2016-09-30 | PaddlePaddle | Baidu | library | 23.94K | |||||||||
| 2016-04-27 | OpenAI Gym | OpenAI | library | 37.22K | |||||||||
| 2016-01-27 | ★ AlphaGo | paper | |||||||||||
| 2015-11-09 | ★ TensorFlow | library | 195.63K | 8.82K | |||||||||
| 2015-03-09 | ★ Distilling the Knowledge in a Neural Network | paper | 13.96K | ||||||||||
| 2015-02-26 | ★ DQN (Deep Q-Network) | paper | |||||||||||
| 2015-02-11 | ★ Batch Normalization | paper | 24.38K | ||||||||||
| 2014-12-17 | Deep Speech 1 & 2 | Baidu | paper |