Moonshot releases Kimi K3.1 Pro, which it says scores 65.7% on Terminal-Bench 4.0, just behind Claude Opus 5.5, and promises the full weights by 4 November

Moonshot AI released Kimi K3.1 Pro, the first model of its K3.1 series, in its apps and API. Its blog calls it “our most capable 3T-class open-weight model” and gives no exact parameter count. The model has a 1-million-token context. A faster variant, K3.1 Pro Highspeed, is also available. Moonshot says the full model weights will be released by 4 November, so the model is not yet open.

On Moonshot's own figures, K3.1 Pro scores 65.7% on Terminal-Bench 4.0, against 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra. On Moonshot's internal hallucination test, which it calls a stress test, the claim-level hallucination rate falls from 23.4% for K3 to 8.3%. Moonshot says it removed cyber-related data from training ahead of the open-weight release.

Artificial Analysis, which measures models independently, gives K3.1 Pro 55 on its Intelligence Index. That is fourth place, 11 points above Kimi K3, and behind Claude Opus 5.5 at 58 and Claude Sonnet 5.5 and Muse Wonder Full at 56. It lists the price as $2 per million input tokens and $10 per million output tokens, a third below K3.

The Moonshot page carries no visible date; Artificial Analysis and The Decoder reported the launch on 23 October.

• Claimed K3.1 Pro scores 65.7% on Terminal-Bench 4.0, against 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra, Moonshot says. Moonshot AI, Kimi blog
• Claimed The full model weights will be released by 4 November, Moonshot says. Moonshot AI, Kimi blog
• Confirmed Artificial Analysis scores it 55 on its Intelligence Index, fourth, at $2 input and $10 output per million tokens. Artificial Analysis