Kimi K3 is a 2.8T-parameter model with Kimi Delta Attention and Attention Residuals, native vision, 1M context. Available on Kimi.com, Work, Code, API. Full model weights will be released by July 27, 2026. MXFP4 weights with MXFP8 activations via quantization-aware training. Recommend deploying on supernode configurations with 64 or more accelerators. MoE: 16 of 896 experts. Pricing 3.00/$15.00 per MTok cache-hit/miss/output.