Moonshot AI introduced Kimi K3.1 Pro on 2026-10-24: the first model in its new K3.1 series and, per the company, its most capable 3T-class open-weight model to date, with a 1-million-token context window. Official messaging highlights gains in coding, 3D creation, financial research, and other long-horizon agentic work. A companion Kimi K3.1 Pro Highspeed variant offers the same intelligence at roughly 2–3× output speed. Both are available now on kimi.ai / the Kimi apps, Kimi Work, Kimi Code, and the Kimi API; full model weights are slated for release by 4 November 2026.
Post-trained K3: 65.7% on Terminal-Bench 4.0
K3.1 Pro is a post-trained version of Kimi K3 with substantially larger post-training compute. Moonshot says it rebuilt layers of the stack—RL framework, long-context inference, training efficiency, sandboxes, and scheduling—so large-scale joint RL across hundreds of domains yields stable gains.
On Terminal-Bench 4.0 (realistic, resource-constrained terminal tasks with strict task-specific tests), K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%). On the internal Kimi Code Bench 3.0 Long-horizon (multi-step execution, tool use, end-to-end engineering), it scores 61.2%, above K3’s 55.95%.
Coding, 3D, and finance workflows
Translucent blueprint cubes, a parametric mechanical form, and financial chart ribbons orbiting a glowing central prism — coding, 3D creation, and financial-research agents (editorial illustration; no text or logos).
Moonshot positions K3.1 Pro for the full coding range—from fixes in real repositories to full-project builds—and for game engines such as Unity, Unreal Engine, and Godot, combined with web search and 3D generation. On BenchCAD (3D reconstruction from multi-view renderings via CAD code), with tools it scores 92.7%, approaching Claude Opus 5.5 (96.2%) and GPT-6 Astra (95.9%). 3D Make is live on kimi.ai and the Kimi app for online preview/edit, download for 3D printing, or export as video / Live Photos.
In finance, OfficeQA Pro v2 (retrieval, cross-document checks, multi-step numerical reasoning over real financial documents) scores 64.4, which Moonshot calls state of the art. Internal finance benches: Financial Modeling +7.2 points vs K3, Financial Research +11.6 points, Business Operation about 4.2× K3. On the internal Kimi Hallu Bench Hard stress test, claim-level hallucination fell from 23.4% (K3) to 8.3%, and responses with at least one factual error were about 42.4% of the prior rate (Moonshot notes this is a deliberately hard stress set, not everyday error rates).
Cyber controls ahead of open weights
A dedicated cyber section says Moonshot identified and removed cyber-related data from mid-training and RL environments for the open-weight release, while keeping general-purpose coding tasks. On internal xdev-bench-hard (source-code vulnerability discovery/exploitation, pass@3), a checkpoint without those risk-reduction steps scored 90.5%; final K3.1 Pro scores 81.0%. The company says offensive cyber capability remains behind frontier proprietary models that have been public for at least three months, and that it aims not to release open weights that would extend the absolute cyber frontier.
Availability, API pricing, and reasoning settings
Channels: latest Kimi app (iOS, Android, HarmonyOS) or kimi.ai; Kimi Work desktop 3.1.0+ (Windows and Apple-silicon Macs); Kimi Code via /model. Enterprise access is off by default until an admin enables it.
API model IDs: kimi-k3.1-pro and kimi-k3.1-pro-highspeed; context window 1,048,576 tokens with flat pay-as-you-go pricing (no context-length tiers). Platform list prices per million tokens (USD, excl. tax): standard $2.00 input / $10.00 output / $0.20 cache hit / cache write $2.00 (5m TTL) or $4.00 (1h TTL); Highspeed $4.00 input / $20.00 output / $0.20 cache hit / write $4.00 (5m) or $8.00 (1h). Flagship unlock requires a successful top-up (English docs: minimum $1). Thinking is always on; reasoning_effort supports low / high / max (default max). Image and video inputs and tool calling are supported; Zero Data Retention is available for API customers.
Primary sources:
- Moonshot / Kimi, “Kimi K3.1 Pro：智能前沿，更进一步,” 2026-10-24, https://www.kimi.com/news/kimi-k3-1-pro
- Moonshot / Kimi, “Kimi K3.1 Pro Tech Blog: Advancing Open Intelligence,” 2026-10-24, https://www.kimi.ai/blog/kimi-k3-1-pro
- Kimi API Platform, “Kimi K3.1 Pro” quickstart and pricing, https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart ; https://platform.kimi.ai/
