Kimi K3.1 Pro nearly ties Opus at coding, but you can't download it yet
Kimi K3.1 Pro, by Moonshot AI, has 1M context and 65.7% on Terminal-Bench 4.0. Open weights promised by November 4.
Kimi K3.1 Pro is Moonshot AI's new model, launched on October 23, 2026, with a one-million-token context window. According to the company itself, it nearly ties Claude Opus at coding: it scores 65.7% on Terminal-Bench 4.0, just under Opus. The detail almost nobody mentions: that comparison comes from Moonshot, and the open weights aren't published yet; the company promised to release them by November 4.
What Kimi K3.1 Pro is
An AI that lives in someone else's cloud is never yours. That's why the promise to open the weights matters. Kimi K3.1 Pro reads one million tokens at once and, according to Moonshot, codes whole projects, models 3D parts and analyzes finance.
What it does
- It codes whole projects.
- It models 3D parts.
- It analyzes finance.
- It reads one million tokens of context in a single pass.
On Terminal-Bench 4.0 it scores 65.7%, just under Opus, according to Moonshot. That is a manufacturer figure, not an independent test.
Why you can't download it yet
The open weights aren't published. Moonshot promised to release them by November 4. Meanwhile it only runs through Kimi.com, Kimi Work, Kimi Code and Moonshot's API. The license hasn't been published either, and we couldn't confirm per-token pricing, so we don't give it.
How to apply it in your business?
- Try it today through Kimi.com, Kimi Work, Kimi Code or the API if you have a coding task you currently pay another cloud for.
- Measure it yourself: the comparison with Opus is Moonshot's; validate it on your own cases.
- Wait for November 4 if being able to choose where to run it matters to you: from then on, your team could decide.
- Check the license once it's published before using it in a product.
If you're interested in open-weights models and their pricing, read the full breakdown of DeepSeek V4.1 Flash.
Frequently asked questions
What is Kimi K3.1 Pro?
Moonshot AI's model launched on October 23, 2026, with a one-million-token context window.
Can I download it?
Not yet. Moonshot promised to publish the open weights by November 4.
Where does it run today?
On Kimi.com, Kimi Work, Kimi Code and Moonshot's API.
Does it really nearly tie Opus?
That's what Moonshot says: 65.7% on Terminal-Bench 4.0, just under Opus. It's the company's own comparison.
How much does it cost and what is the license?
The license hasn't been published and we couldn't confirm per-token pricing.
Conclusion
Kimi K3.1 Pro arrives with a million tokens of context and a near-Opus coding figure, but all of it is Moonshot's and the open weights don't exist yet. Try it in its cloud, don't plan around November 4 until it happens, and check the license when it lands. Sources: Moonshot AI's official blog and SiliconANGLE (October 23, 2026).