Moonshot AI Launches Kimi K3 with 2.8 Trillion Parameters, Full Weights to Open-Source on July 27

According to Beating, Moonshot AI has launched Kimi K3, its flagship large language model featuring 2.8 trillion parameters and support for 1 million token context window. The model employs a mixture-of-experts architecture, activating only 16 out of 896 experts per inference. Moonshot claims this is the first 3-trillion-scale open model.

In internal benchmarks, K3 outperforms Claude Opus 4.8 and GPT-5.5 on select coding and Agent tasks, though it trails Claude Fable 5 and GPT-5.6 Sol. K3 is available via Kimi, Kimi Work, Kimi Code, and API, priced at $3 per million input tokens and $15 per million output tokens. Full model weights are scheduled for open-source release by July 27.