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

Kimi K3 has 2.8 trillion parameters, 1 million token context, and native multimodal capabilities. It features Kimi Delta Attention for up to 6.3x faster decoding and Attention Residuals for ~25% higher training efficiency. On benchmarks, it scores comparably to Opus 4.8 and GPT-5 on the Artificial Analysis Intelligence Index.