Kimi K3.1 Pro: Specs, Pricing and What It Means for Small Teams

Key takeaway: Kimi K3.1 Pro is worth a test if you pay for AI by usage. Moonshot AI launched it in the past few days as its most capable open-weight model so far, with a 1-million-token context window. On the API it costs $2 per million input tokens and $10 per million output tokens (USD, before tax), which is less than the K3 model it replaces. The benchmark results are Moonshot’s own, and the open weights aren’t out yet.

• Price: pay as you go, no subscription. A faster Highspeed version costs double. (Kimi API pricing)
• Dates: full weights promised “by Nov 4, 2026”. K2 models close to new users on 31 October.

Research-based news. We haven’t tested Kimi K3.1 Pro. Facts come from Moonshot’s and Anthropic’s own pages, checked 28 October 2026. Prices are in USD and exclude tax.

What launched

Moonshot AI announced Kimi K3.1 Pro in the past few days. Key points from the launch post:

• What it is: the first K3.1 model, a post-trained version of K3 that Moonshot calls its “most capable 3T-class open-weight model to date”.
• Context window: 1 million tokens (1,048,576 on the API), enough for several long reports in one request.
• Highspeed version: Moonshot says K3.1 Pro Highspeed runs 2 to 3 times faster.
• Open weights: “Full model weights will be released by Nov 4, 2026.” The licence terms aren’t published yet, so don’t assume commercial use is free.

Kimi K3.1 Pro pricing

From the Kimi API pricing page, USD per million tokens, checked 28 October 2026:

Model	Input	Output	Cached input	Cache write (5 min)	Cache write (1 hour)
kimi-k3.1-pro	$2.00	$10.00	$0.20	$2.00	$4.00
kimi-k3.1-pro-highspeed	$4.00	$20.00	$0.20	$4.00	$8.00
kimi-k3 (previous)	$3.00	$15.00	$0.30	$3.00	$6.00

• Caching is automatic (default 5-minute window), so a resent long prompt, like a price list, costs $0.20 per million tokens for the cached part. File APIs are temporarily free.
• Example, not a quote: 20,000 tokens in and 2,000 out comes to about (0.02 × $2) + (0.002 × $10) = $0.06, before tax and without caching.

Kimi app prices: the subscription prices for Kimi.ai, Kimi Work and Kimi Business, and whether the free tier includes K3.1 Pro, are unconfirmed.

How it compares on price

Anthropic’s pricing page lists Claude Opus 5.5 at $4 input and $20 output per million tokens, and Claude Sonnet 5.5 at $2 and $10. So K3.1 Pro’s list price is half of Opus 5.5’s and the same as Sonnet 5.5’s. Quality and tokens used per task also change the real cost. If you just want a chat app, start with our ChatGPT vs Claude vs Gemini comparison.

What Moonshot says it’s better at

All figures below are Moonshot’s own reported results. We haven’t tested them.

Test (Moonshot’s description)	K3.1 Pro	Compared with
Terminal-Bench 4.0 (coding in a terminal)	65.7%	Claude Opus 5.5 66.4%; GPT-6 Astra 58.2%
Kimi Code Bench 3.0, long tasks (internal)	61.2%	Kimi K3 55.95%
BenchCAD, with tools	92.7%	Opus 5.5 96.2%; GPT-6 Astra 95.9%
OfficeQA Pro v2	64.4	Moonshot calls it state of the art

• Finance tasks (Moonshot’s internal tests): Financial Modeling up 7.2 points and Financial Research up 11.6 points on K3; Business Operation about 4.2 times K3.
• Hallucinations: on Moonshot’s internal stress test, Hallu Bench Hard, the claim-level error rate fell from 23.4% to 8.3%. That’s a deliberately hard test, not an everyday error rate.

Where you can use it

• Kimi.ai, the Kimi apps (iOS, Android, HarmonyOS), Kimi Work desktop (v3.1.0+) and Kimi Code (/model).
• The Kimi API: model names kimi-k3.1-pro and kimi-k3.1-pro-highspeed. Zero Data Retention is available to API customers.
• Kimi Enterprise: admins can switch it on; it’s off by default at launch.
• 3D Make, for makers and creators: live on kimi.ai and in the app. It exports models for 3D printing and works with Unity, Unreal Engine and Godot.

Dates to know

• 31 October 2026: K2 models close to new users.
• By 4 November 2026: full K3.1 Pro weights due.
• End of December 2026: all K2 models retire. If you build on K2, plan the switch now.

Should a small team try it?

Try it on one real task, like summarising long documents, and compare cost per finished task, not price per token. Before sending customer data, check Kimi’s privacy terms for where data is stored, and use Zero Data Retention on the API. If you call models from workflow tools, see our n8n vs Make vs Zapier comparison.

A cheaper model only helps when it does real work. AIFix Automation can put a model like this to work in your booking and follow-up flow, on accounts in your name. That’s our own service.

What we checked, and what we’ll test

We read Moonshot’s launch post and the Kimi and Claude pricing pages on 28 October 2026. We haven’t run the model. Next: confirm the app plans, run one long-document task through the API, and add the licence once the weights ship.

FAQ

What is Kimi K3.1 Pro pricing?

On the API: $2 per million input tokens, $10 per million output tokens and $0.20 per million cached tokens (USD, before tax). Highspeed is $4 and $20. App prices are unconfirmed.

Is Kimi K3.1 Pro better than Claude?

We can’t say. Moonshot’s own results put it near Opus 5.5 on some tests and behind on others, and we haven’t tested it.