Where it comes from
Moonshot AI describes Kimi K3.1 Pro as a post-trained version of Kimi K3, trained with substantially more reinforcement learning across hundreds of task domains. It launched together with Kimi K3.1 Pro Highspeed, which Moonshot says runs 2 to 3 times faster.
At launch it was available in the Kimi apps, Kimi Work, Kimi Code and the Kimi API, and on OpenRouter as moonshotai/kimi-k3.1-pro.
What it is good at
Moonshot's announcement highlights four areas. The numbers below are Moonshot's own evaluations, not independent results.
- Coding: 65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra (58.2%) and just behind Claude Opus 5.5 (66.4%).
- 3D and design: 92.7% on BenchCAD with tools, with Blender and CAD scripting.
- Finance and research: 64.4 on OfficeQA Pro v2, which Moonshot calls state of the art.
- Factual accuracy: the claim-level hallucination rate on Moonshot's internal stress test fell from 23.4% for K3 to 8.3%.
How to try it
You can use Kimi K3.1 Pro in the official Kimi apps, through the Kimi or OpenRouter APIs, or in an independent web chat such as this one.
- Open the chat. Sign in on kimi-k31.com and type a question, paste a long document, or attach images or a short video.
- Pick a speed. K3.1 Pro by default, or Highspeed when you want faster answers at twice the token price.
- Read and follow up. The reasoning streams first, then the answer; follow up in the same thread.