K2 Think V2

70B dense general-reasoning model from MBZUAI's Institute of Foundation Models, post-trained from K2-V2 Instruct via two-stage RLVR (Reinforcement Learning with Verifiable Rewards) on the improved Guru dataset. Released January 27, 2026 as the UAE's fully sovereign reasoning system — pre-training, post-training, weights, and reasoning data all open under Apache 2.0.

AA Intelligence Index 24 (well above the 15 median for similar-size open-weight models) at a 262K-token context window. Improves on K2-V2 by 4 points on the Intelligence Index, with hallucination rate dropping from 89% to 52% and long-context reasoning from 33% to 53%. Tied for first place with K2-V2 and Olmo 3 on Artificial Analysis's Openness Index.