Kimi K3.1 Pro: 2.8T open model trails only Opus 5.5

Moonshot's Kimi K3.1 Pro is out. It's a 2.8T-parameter MoE with 104B active per token, a 1M-token context window and text+image input. Moonshot says full open weights land by November 4.

The "on par with Opus 5.5" claim holds up on some tests and not others. Moonshot's own numbers put it at 65.7% on Terminal-Bench 4.0, against 66.4% for Opus 5.5 and 58.2% for GPT-6 Astra. On BenchCAD with tools it trails both (92.7% vs 96.2% and 95.9%). Artificial Analysis ranks it 4th on its Intelligence Index with 55, behind Opus 5.5 (58), Sonnet 5.5 (56) and Muse Wonder Full (56). Most of the benchmark figures are Moonshot's own.

It's cheaper than the closed models it chases. Pricing is $2 per 1M input and $10 per 1M output tokens, 33% below K3. Moonshot also says it removed cyber-related data from both training stages ahead of the open release.

Why it matters: once the weights ship, Artificial Analysis expects this to be the strongest open-weights model, a few points behind the best closed ones at a lower price.