OpenAI

startup

usa · Founded 2015 · Valuation: $852B as of 2026-03

OpenAI was founded in December 2015 as a nonprofit AI research lab by Sam Altman, Ilya Sutskever, Greg Brockman, and others. It restructured to a capped-profit entity in 2019 and completed conversion to a Public Benefit Corporation in October 2025, with Microsoft holding ~27% equity. In March 2026 OpenAI raised $122B at an $852B valuation — the largest private funding round in history — from SoftBank, Amazon, NVIDIA, and others. Revenue reached ~$2B/month by early 2026 with 900M+ weekly ChatGPT users.

OpenAI's GPT series defined the modern LLM era: GPT-2 (2019, 1.5B) demonstrated emergent text generation, GPT-3 (2020, 175B) introduced few-shot learning, and GPT-4 (2023) set the frontier for multimodal reasoning. The o-series reasoning models (o1, o3) pioneered test-time compute scaling, and GPT-5.4 (2026, AA index 57, 1M context) is the current flagship. Beyond language, OpenAI created DALL-E (text-to-image), Whisper (speech recognition), Codex (powering GitHub Copilot), and Sora (video generation).

The lab's foundational technique papers reshaped the entire field: PPO (2017) became the default RL algorithm for alignment, Scaling Laws (2020) established power-law relationships driving compute investment, CLIP (2021) enabled text-image alignment used by virtually all image generation models, and InstructGPT/RLHF (2022) created the SFT → Reward Model → PPO pipeline now standard across the industry. OpenAI also released influential open-source tools including Gym (37K stars) and Triton (19K stars).

Website Wikipedia HuggingFace GitHub Artificial Analysis OpenRouter
frontierreasoningmultimodalcodingagents

People

News

Outputs (21)

GPT-5.4

model

Current flagship model. 1M token context. Knowledge cutoff August 31, 2025. Available in thinking, pro, mini, and nano variants. Architecture and para…

GPT-5

model

Multi-component system combining a fast model, reasoning model, and router into a single interface. 400K token context. Replaced GPT-4o, o3, o4-mini,…

o3

model

Second-generation reasoning model, advancing o1's test-time compute scaling. 200K token context. Released alongside o4-mini, a smaller reasoning varia…

o1

model

OpenAI's first reasoning model, trained with reinforcement learning to perform extended chain-of-thought reasoning before answering. Uses hidden "thin…

GPT-4o

model

GPT-4o ("omni") — OpenAI's first natively multimodal model, processing text, audio, image, and video inputs and generating text and audio output…

Sora

model

Text-to-video generation model using a diffusion transformer operating on spacetime patches. Generates up to 1080p video, up to 20 seconds long, from…

DALL-E 3

model

"Improving Image Generation with Better Captions" — the third generation of DALL-E, integrated natively into ChatGPT. Key innovation is a captio…

Let's Verify Step by Step

paper

Demonstrated that process reward models (PRMs) that evaluate each step of a reasoning chain significantly outperform outcome reward models (ORMs) that…

GPT-4

model

GPT-4 technical report. Architecture details undisclosed by OpenAI (parameter count, training data, and hardware not published). 8K token context at l…

Whisper

model

"Robust Speech Recognition via Large-Scale Weak Supervision" — an encoder-decoder Transformer trained on 680,000 hours of multilingual audio dat…

InstructGPT (RLHF)

paper

"Training language models to follow instructions with human feedback" — introduced the three-stage RLHF pipeline that became industry standard:…

Codex

model

"Evaluating Large Language Models Trained on Code" — GPT-3 fine-tuned on publicly available code from GitHub. The largest variant, Codex-12B, so…

Triton

library

Open-source GPU programming language and compiler for writing high-performance neural network kernels. Python-like syntax that compiles to optimized G…

CLIP

paper

"Learning Transferable Visual Models From Natural Language Supervision" — Contrastive Language-Image Pre-training (CLIP) trained on 400M image-t…

DALL-E

model

"Zero-Shot Text-to-Image Generation" — a 12B parameter autoregressive Transformer that generates images from text descriptions. Uses a discrete…

GPT-3

model

"Language Models are Few-Shot Learners" — 175B parameter dense Transformer (96 layers, 12288 hidden, 96 heads) trained on 300B tokens from a fil…

Scaling Laws for Neural Language Models

paper

Established power-law scaling relationships between language model performance and three key factors: model size (N), dataset size (D), and compute bu…

GPT-2

model

"Language Models are Unsupervised Multitask Learners" — 1.5B parameter decoder-only Transformer (48 layers, 1600 hidden, 25 heads) trained on We…

GPT-1

model

"Improving Language Understanding by Generative Pre-Training" — the paper that launched the GPT series. A 117M parameter decoder-only Transforme…

Proximal Policy Optimization (PPO)

paper

Introduced Proximal Policy Optimization, a family of policy gradient methods that alternate between sampling data through interaction with the environ…

OpenAI Gym

library

Toolkit for developing and comparing reinforcement learning algorithms. Provides a standard API for RL environments (CartPole, Atari, MuJoCo, etc.) th…