Moonshot AI releases Kimi K3, a 2.8T parameter open-weight model

Kimi K3 is a 2.8T parameter sparse Mixture-of-Experts model featuring native vision support and a 1M token context window. It achieves 89.4% pass@4 on the DeepSWE benchmark, outperforming proprietary systems like GPT-5.6 Sol in multi-attempt coding tasks while costing 64% less per rollout.