What Is Kimi K3.1 Pro? Price, Highspeed and Open Weights
The short answer
Kimi K3.1 Pro is Moonshot AI’s new flagship, released October 23, 2026: a post-trained Kimi K3 with a 1M-token context, priced at $2 input / $10 output per million tokens — a third cheaper than K3 — with open weights promised by November 4, 2026. A faster sibling, K3.1 Pro Highspeed, costs $4 / $20.
Key facts (as of October 25, 2026)
| Kimi K3.1 Pro | K3.1 Pro Highspeed | Kimi K3 (previous) | |
|---|---|---|---|
| API ID | kimi-k3.1-pro | kimi-k3.1-pro-highspeed | kimi-k3 |
| Input / output per MTok | $2.00 / $10.00 | $4.00 / $20.00 | $3.00 / $15.00 |
| Cache hit | $0.20 | $0.20 | $0.30 |
| Cache write (5 min / 1 h) | $2.00 / $4.00 | $4.00 / $8.00 | $3.00 / $6.00 |
| Context | 1,048,576 tokens | 1,048,576 tokens | 1,048,576 tokens |
| Speed | Baseline | 2–3× (Moonshot) | — |
| Weights | Due by Nov 4, 2026 | — | Open since July 27, 2026 |
Prices from Moonshot’s pricing page, which states K3-series pricing is flat — no tiering by context length.
What changed from Kimi K3
Moonshot’s model guide calls K3.1 Pro “a post-trained version of Kimi K3, with substantially larger post-training compute”, trained with joint reinforcement learning across hundreds of domains, and says it improves coding, 3D creation, financial research and agentic work. Its showcase emphasises full-project builds in game engines (Unity, Unreal Engine, Godot) and polished front-end and Three.js work. Moonshot calls it its “most capable 3T-class open-weight model”; press coverage puts it at roughly 2.8 trillion total parameters with about 104 billion active (a mixture-of-experts design), which Moonshot’s own guide does not state.
On benchmarks, be careful. On October 25, 2026, Moonshot’s English model guide still showed placeholder brackets (“[xx.x%]”) where its Terminal-Bench 4.0 figures should be, and third-party write-ups quote different numbers for the same test. Wait for independent evaluations before treating any single score as settled.
How to use it
- API: OpenAI-compatible endpoint on platform.kimi.ai. K3.1 Pro is a “flagship” model, unlocked after a minimum $1 top-up; your cumulative top-up sets your rate-limit tier. Context caching is automatic.
- OpenRouter:
moonshotai/kimi-k3.1-proat $2 / $10 with $0.20 cached input; not available through OpenRouter’s EU in-region routing as of October 25. - Apps: Kimi web and mobile apps, Kimi Work and Kimi Code.
Migration note: Moonshot’s pricing page says K2-series models stop being available to new users on October 31, 2026 and will be retired by the end of December 2026 — move K2 workloads to K3 or K3.1 Pro.
Is it worth switching?
- From Kimi K3 — yes for most workloads: same context, a third cheaper on every line, and more post-training.
- From Claude Sonnet 5.5 — same $2/$10 headline price; test on your own tasks, and weigh open weights (from November 4) against Anthropic’s tooling and cloud availability.
- For self-hosting — wait for the weights and the licence; a 3T-class model needs multi-node GPU capacity.
Compare with other open-weight options in DeepSeek V4 Pro vs GLM-5.3 vs Kimi K3 and the full current API prices table.
Last verified: October 25, 2026.