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

Kimi K3 has 2.8 trillion parameters, 1 million token context, and native multimodal capabilities with Kimi Delta Attention for 6.3x faster decoding. Compared to Claude Fable 5 on DeepSWE, K3 achieves near-flagship coding performance at ~35% cost and pulls ahead at higher pass@k.