Kimi K3

Kimi K3 is Moonshot AI's flagship model with a 2.8-trillion-parameter sparse MoE (16 of 896 experts active per token), a 1M-token context, native vision, and multi-agent support, featuring Kimi Delta Attention for faster long-context decoding. Priced at $0.30 cache-hit, $3.00 cache-miss input, and $15.00 output per million tokens, it always reasons with a reasoning_effort field (low/high/max) and K3-only tool constraints; open weights (2.8T total/104B active, 93 layers, MoonViT-V2 encoder, MXFP4/MXFP8 quantized) are available on Hugging Face under a permissive license, with a separate agreement required for Model-as-a-Service revenue over $20M annually.