Moonshot AI has introduced Kimi K3.1 Pro, the first model in its new K3.1 series and its most capable 3T-class open-weight model to date, featuring a 1-million-token context window. The company says it delivers substantial improvements across coding, 3D creation, financial research and other agentic capabilities. It is also releasing Kimi K3.1 Pro Highspeed, which runs at 2-3× the speed of K3.1 Pro for more responsive interactions. Both models are available today on Kimi.ai, Kimi Work, Kimi Code and the Kimi API, with full model weights due by Nov 4.
K3.1 Pro is a post-trained version of Kimi K3, built with substantially larger post-training compute. The team optimized the full stack, including the RL framework, long-context inference and sandbox infrastructure, so that each increment of RL compute delivers stable, predictable returns. On Terminal-Bench 4.0, K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%). On the company's internal Kimi Code Bench 3.0 Long-horizon test, it scores 61.2%, above K3's 55.95%.
For coding, the model handles everything from quick fixes to full-project builds and works across Unity, Unreal Engine and Godot. In one internal test, it built a human walking dynamics simulator, moving the simulated figure from collapsing within seconds to walking continuously for 11.2 seconds after 167 missions.
The model is also strong at 3D creation. On Bench CAD it scores 92.7% with tools, approaching Claude Opus 5.5 (96.2%) and GPT-6 Astra (95.9%), and it excels at parametric modeling of mechanical parts, architectural models and custom furniture.
Source: Hacker News · Summarized by HeadlinesBriefing