Kimi launches K3.1 Pro and K3.1 Pro Highspeed

Moonshot AI has released Kimi K3.1 Pro, a post-trained upgrade that scores 55 on the Artificial Analysis Intelligence Index, up 11 points from K3, placing it fourth behind Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. If weights are released as planned by November 4, 2026, it will become the highest-scoring available open-weights model and narrow the open-vs-proprietary gap to three points. A faster K3.1 Pro Highspeed variant is also available now.

Moonshot has released Kimi K3.1 Pro, an upgraded version of its open-weight Kimi K3, with full weights due by 4 November. The lab says the model nearly matches Claude Opus 5.5 on a coding test and that it removed hacking-related training data to limit misuse of the open release.

Kimi unveiled K3.1 Pro, a 3T-class open-weight model with a 1M-token context window, plus K3.1 Pro Highspeed, with full weights to be released by November 4; it also lists multiple benchmarks.

According to TestingCatalog relaying Kimi AI, Kimi K3.1 Pro is a 2.8T total / 104B active parameter MoE open model that performs on par with leading models like Opus 5.5 and GPT-6 Astra at a significantly lower price, with open weights coming soon.

Artificial Analysis has released the full Intelligence Index results for Kimi K3.1 Pro, which ranks fourth overall with a score of 55. The model shows major gains over Kimi K3 in agentic tasks and sets a new bar for models of comparable scale.

Kimi K3.1 Pro first-party pricing is $2 per million input tokens, $10 per million output and $0.20 per million cached input, 33% below Kimi K3. Artificial Analysis estimates its cost at $1.98 per Intelligence Index task versus $2.00 for Kimi K3.

Kimi K3.1 Pro outputs 55K tokens per Intelligence Index task versus 48K for Kimi K3, up 14%. Output rises on agentic evaluations such as AutomationBench-AA and GDPval-AA while falling by more than half on AA-Omniscience.

Kimi K3.1 Pro keeps 2.8T total and 104B active parameters, the same size as Kimi K3, yet scores 11 points higher on the Intelligence Index. Artificial Analysis places it at the top of the Pareto frontier against both total and active parameters.

Kimi K3.1 Pro's AA-Omniscience Index rises to 30 from 20 for Kimi K3. Accuracy moves slightly from 48% to 51% while the hallucination rate falls from 53% to 44%, driving most of the index gain.

Kimi K3.1 Pro scores 1,846 Elo on GDPval-AA, third overall and up 312 Elo from Kimi K3, behind only Claude Opus 5.5 and Muse Wonder Full. Artificial Analysis reports it also achieves the highest score of any model on Terminal-Bench 4.0.

Kimi K3.1 Pro launches at fourth on the Artificial Analysis Intelligence Index with 55 points, up 11 points from Kimi K3, with large gains in agentic performance. Its weights have not yet been published; once released it should be the leading open-weights model.

Moonshot AI has released Kimi K3.1 Pro, drawing 58 points on Hacker News. Only the name and early attention are public so far; no capability, parameter, or pricing details are given.

Kimi K3.1 Pro is rolling out on Kimi AI web and mobile, positioned as the flagship model for complex tasks and professional work. TestingCatalog reports the rollout has begun, so users can check the web app or update their app to see if the version is available yet.