Kimi K3.1 Pro is Moonshot AI’s newest flagship model, and it lands less than a point behind Claude Opus 5.5 on Terminal-Bench 4.0 at about a fifth of the output price. Moonshot calls it its “most capable 3T-class open-weight model”, and the full weights are due by 4 November 2026. In the video below I put it side by side with Claude on real builds. Here’s what it is, how it scores, what it costs and where it still falls short.
Key takeaways
- Kimi K3.1 Pro is a post-trained version of Kimi K3 with far more post-training compute.
- 65.7 on Terminal-Bench 4.0 vs 66.4 for Opus 5.5 and 58.2 for GPT-6 Astra.
- API price: $2 in / $10 out per million tokens, with a 1M-token context.
- Open weights promised by 4 November 2026. Claim a free strategy session to plan your model stack.
What is Kimi K3.1 Pro?
Kimi K3.1 Pro is the first model in Moonshot’s K3.1 series. According to the official Kimi blog, it isn’t a brand-new base model. It’s Kimi K3 with a much bigger post-training run on top. That’s where the gains come from: better long coding jobs, better tool use and fewer made-up claims.
Moonshot describes it as a “3T-class” open-weight model. It hasn’t published an exact parameter count for K3.1 Pro, so I won’t guess one. What we do know is the context window: 1 million tokens. That’s enough for a big codebase, a long contract pack or months of notes in one go.
There’s also a Kimi K3.1 Pro Highspeed variant. It’s the same model served faster — Moonshot says two to three times the speed — for twice the price.
| Kimi K3.1 Pro at a glance | |
|---|---|
| Maker | Moonshot AI |
| Base | Post-trained Kimi K3 |
| Size | “3T-class” (no exact count published) |
| Context | 1M tokens |
| Variants | Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) |
| Open weights | Full weights by 4 November 2026 |
Kimi K3.1 Pro benchmarks
Here are the numbers Moonshot published. Remember these are the maker’s own results, so treat them as a starting point and test on your own work.
| Benchmark | Kimi K3.1 Pro | Claude Opus 5.5 | GPT-6 Astra |
|---|---|---|---|
| Terminal-Bench 4.0 | 65.7 | 66.4 | 58.2 |
| BenchCAD (with tools) | 92.7 | 96.2 | 95.9 |
| BenchCAD (no tools) | 72.8 | 73.0 | — |
The headline is Terminal-Bench. That test checks whether a model can actually get work done in a terminal, step by step. K3.1 Pro sits 0.7 points behind Opus 5.5 and well ahead of GPT-6 Astra. On BenchCAD without tools it’s almost level with Opus 5.5 and well clear of Fable 5.1 (60.6).
On long coding jobs, Moonshot’s internal Kimi Code Bench 3.0 went from 55.95 on K3 to 61.2 on K3.1 Pro. Business work improved too: OfficeQA Pro v2 is 64.4, which Moonshot claims is the best score so far, and Frontier Finance is 58.6. The biggest jump was in an internal business-operations test, from 8.9 to 37.
The demo that stuck with me is the walking test. Moonshot gave it a human body model with 296 skeleton parts and 402 muscle parts. The walk went from collapsing straight away to 11.2 seconds of steady walking, after the model found and fixed 36 shoulder-muscle errors.
Want Kimi K3.1 Pro working inside your business? Inside the AI Profit Boardroom you get the Agent OS — Hermes, OpenClaw, OpenCode, Claude and Codex in one place — plus step-by-step Hermes training and four live coaching calls a week.
Want a plan for your business first? Claim your free strategy session.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →
Kimi K3.1 Pro pricing
This is where it gets interesting. Here are the prices from the Kimi API pricing page, per million tokens:
| Model | Input | Output | Cached input |
|---|---|---|---|
| kimi-k3.1-pro | $2.00 | $10.00 | $0.20 |
| kimi-k3.1-pro-highspeed | $4.00 | $20.00 | $0.20 |
| kimi-k3 (previous) | $3.00 | $15.00 | $0.30 |
So K3.1 Pro is actually cheaper than K3 while scoring higher. Cache writes cost $2.00 for a 5-minute cache and $4.00 for a 1-hour cache. Caching is automatic, which helps a lot if you keep sending the same big codebase or system prompt.
One heads-up: the K2 series closes to new users on 31 October 2026 and retires at the end of December 2026. If you’re still on K2, now is a good time to move.
Want help picking the right AI model for your business? Claim your free strategy session →
Where you can use Kimi K3.1 Pro
- Kimi.ai on the web, plus the iOS, Android and HarmonyOS apps.
- Kimi Work desktop app, version 3.1.0 or later, on Windows and Apple silicon Macs.
- Kimi Code — switch with the /model command.
- The Kimi API as
kimi-k3.1-proandkimi-k3.1-pro-highspeed, with zero data retention available. - Business and Enterprise workspaces, where an admin has to switch it on first (it’s off by default).
The API is OpenAI-compatible, so most tools that talk to OpenAI can talk to Kimi by changing the base URL. You unlock it after a top-up of at least $1. If you want to run it inside Hermes Agent, I wrote a separate step-by-step guide: Kimi K3.1 Pro in Hermes.
Want the short version? I also wrote my Kimi K3.1 Pro breakdown on the AI Profit Boardroom blog.
What I like about it
Three things stand out. First, the price-to-score ratio. Getting within a point of Opus 5.5 on terminal work for $10 per million output tokens is a big deal if you run agents all day. Second, the 1M context with automatic caching, which keeps long sessions cheap. Third, the open weights. Once they land, you can run it on your own hardware or with a host you trust.
It also works well for 3D and game work. Moonshot says it works with Unity, Unreal, Godot and Three.js scenes, and it powers the “3D Make” feature on kimi.ai.
The limits you should know
It isn’t perfect, and Moonshot is upfront about that.
- Hallucinations: on Moonshot’s hard hallucination test with tools, K3.1 Pro made false claims at a rate of 8.3. That’s much better than K3 (23.4) and Opus 5.5 (15.0), but GPT-6 Astra was lower still at 5.5. Moonshot says this is a stress test, not a normal daily rate.
- Cyber security work: Moonshot says K3.1 Pro is behind the top closed models here, and it removed cyber data for the open-weight release.
- Web search tool: the API’s built-in search is being updated, and Moonshot doesn’t recommend it for production right now.
- Thinking is always on. You can lower the reasoning effort for speed, but you can’t switch thinking off.
Check facts in anything customer-facing. Lower hallucination rates are still not zero.
Kimi K3.1 Pro vs Claude Opus 5.5
In my video I gave both models the same build tasks. Opus 5.5 still edges it on the hardest work, and the benchmarks agree. But the gap is small, and Kimi costs a fraction of the price. For most everyday agent jobs — coding, research, documents — I’d be happy running K3.1 Pro and keeping Opus for the hardest jobs.
My verdict on Kimi K3.1 Pro
Kimi K3.1 Pro is the most capable open-weight model I’ve tested this month. It’s close to Opus 5.5 on terminal work, cheaper than the model it replaces, and open weights are on the way. If you run agents every day and watch your costs, Kimi K3.1 Pro deserves a spot in your stack.
About Julian Goldie
I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.
FAQ: kimi k3.1 pro
What is Kimi K3.1 Pro?
Moonshot AI’s newest flagship model, a post-trained version of Kimi K3 with a 1M-token context and open weights due by 4 November 2026.
How much does Kimi K3.1 Pro cost?
$2 per million input tokens and $10 per million output tokens on the Kimi API, with cached input at $0.20. The Highspeed variant is $4 and $20.
Is Kimi K3.1 Pro better than Claude Opus 5.5?
Not quite. It scores 65.7 on Terminal-Bench 4.0 against 66.4 for Opus 5.5, but it costs much less.
Is Kimi K3.1 Pro open source?
Moonshot says the full open weights will be released by 4 November 2026.
How many parameters does Kimi K3.1 Pro have?
Moonshot only describes it as a 3T-class model and hasn’t published an exact parameter count.
Ready to put Hermes to work in your business? Join the AI Profit Boardroom for the Agent OS, setup training and daily updates.
Or claim your free strategy session and we’ll map out your setup.
Starting from zero? Get the free AI course and community.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →Get the FREE AI Course + Community →
Related reading
Last updated October 2026. This page is a living guide to kimi k3.1 pro — the facts here move fast and it is updated as they do.
