Moonshot AI has released Kimi K3, a 2.8-trillion-parameter open-weights model designed for long-horizon coding and deep reasoning. It is the world's first open-source model in the 3-trillion-parameter class and supports a 1 million token context window.
The architecture features Kimi Delta Attention (KDA) and Attention Residuals to facilitate information flow, alongside a Stable LatentMoE framework that activates 16 of 896 experts. Moonshot claims K3 competes at the level of GPT 5.6 Sol and Claude Fable 5 on coding and agentic benchmarks.
Together AI serves the model with an OpenAI-compatible API, offering three levels of reasoning effort and automatic context caching to reduce input costs.