Kimi K2 tech report just dropped! Quick hits: - MuonClip optimizer: stable + token-efficient pretraining at trillion-parameter scale - 20K+ tools, real & simulated: unlocking scalable agentic data - Joint RL with verifiable + self-critique rubric rewards: alignment that adapts - Ultra-sparse 1T MoE: open-source SoTA on agentic tasks Sharing the path, not just the results — toward open AGI built on transparency and reproducibility. Tech report👉