Moonshot AI ships Kimi K3.1 Pro, closing in on Claude on coding
A post-trained version of Kimi K3 with a 1M-token context, a 2 to 3x faster Highspeed variant, and a price cut to $2/$10 per million tokens. Open weights land by 4 November.
Moonshot AI released Kimi K3.1 Pro on 23 October, a post-trained version of Kimi K3 and, by its own description, its most capable 3T-class open-weight model to date. It is multimodal, keeps Kimi K3’s 1 million token context window, and ships alongside a Highspeed variant that Moonshot says runs 2 to 3 times faster at comparable capability.
Moonshot cites 65.7% on Terminal-Bench 4.0, a coding-agent benchmark, against 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra. It also reports 92.7% on BenchCAD, a 3D-modelling benchmark, and 64.4% on OfficeQA Pro v2, a financial-analysis benchmark. Pricing drops to $2 per million input tokens and $10 per million output tokens, down from $3 and $15 for Kimi K3. Both variants are live now through Moonshot’s API and apps; full weights follow by 4 November.
K3.1 Pro: $2 / $10 per million tokens (input/output)
K3.1 Pro Highspeed: same price, 2-3x the speedWhy it matters: a cheaper, open-weight model claiming to sit within a point of Claude Opus 5.5 on a real coding-agent benchmark is a direct option for anyone running agentic coding workloads at scale and watching the bill.
The caveat: every benchmark number here is Moonshot’s own, and the weights aren’t out yet, so there’s nothing independent to check the coding claim against until 4 November.