Kimi 开源了 K3 模型，Agent Arena 直接干翻 GLM-5.2，净提升超 9%，代码和文本竞技场也是第一，值得关注。
Kimi K3 (Max) 在 Agent Arena 中以 +9.75% 的净提升率超越 GLM-5.2 (Max) 的 +7.12%，成为开源权重模型第一名，并在 Frontend Code 竞技场（1682 分）和 Text 竞技场（1485 分）同样拔得头筹。该模型为 2.8T MoE 架构，支持原生视觉理解与 1M 上下文窗口，号称每单位计算智能提升 2.5 倍。Kimi 团队同步开源模型权重、高性能注意力核、MoE 通信库及代理运行基础设施。
Big update: Among open-weight models, Kimi K3 (Max) is #1 in the Agent Arena with +9.75% net-improve...
Big update: Among open-weight models, Kimi K3 (Max) is #1 in the Agent Arena with +9.75% net-improvement, surpassing GLM-5.2 (Max) at +7.12%, and landed the #1 spot across 5 signals (see below). Kimi K3 (Max) is also now #1 in open-weight in the Frontend Code (1682 pts) and Text (1485 pts) Arenas. Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models get web search, filesystem, and terminal tools to complete complex workflows: writing code, creating slide decks, researching the web, building apps, and analyzing documents. We use causal tracing methodology to measure a model's net improvement, which indicates how much it improves outcomes relative to the average model. Congrats to the @Kimi_Moonshot team for their contribution to the open ecosystem. Kimi.ai @Kimi_Moonshot Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Kim… Tech report: github.com/MoonshotAI/Kim… Tech blog: kimi.com/blog/kimi-k3 🔗 View Quoted Tweet 💬 15 🔄 21 ❤️ 126 👀 10079 📊 24 ⚡