Kimi K3 模型概览：2.8T 参数与 MXFP4 量化
原标题：Kimi K3 Model Overview: 2.8T Parameters, MXFP4 Quantization, and What the Open Weights Mean for the Community
AI 摘要
Moonshot AI 于 2026 年 7 月 16 日发布 Kimi K3，这是首个达到 2.8 万亿参数的开源模型，采用 MoE 架构和 MXFP4 量化，支持 100 万 token 上下文及原生视觉。模型在多个基准测试中表现优异，尤其在编码任务上领先。开源权重计划于 7 月 27 日发布，将推动社区研究和部署。
正文节选
Kimi K3 Model Overview: 2.8T Parameters, MXFP4 Quantization, and What the Open Weights Mean for the Community Moonshot AI publicly released Kimi K3 on July 16, 2026, with full open-source weights promised by July 27. At 2.8 trillion parameters, it is the first open-source model to reach the 3-trillion-parameter class. This article provides a structured model overview for the research and MLOps community. | Field | Value | |---|---| | Developer | Moonshot AI | | Model type | Autoregressive Mixt