Kimi K3.1 Pro: Benchmarks, Price ($2/$10), API & How to Use
Kimi K3.1 Pro launched Oct 23, 2026: $2/$10 per 1M tokens, 1M context, 65.7% on Terminal-Bench 4.0, open weights by Nov 4. Benchmarks, API and how to use it.
Quick answers
- What is Kimi K3.1 Pro?
- Moonshot AI's new flagship model, released Oct 23, 2026. It is a post-trained upgrade of Kimi K3 with a 1M-token context window, built for coding, agents, 3D creation and finance work.
- How much does Kimi K3.1 Pro cost?
- $2 per 1M input tokens, $10 per 1M output tokens and $0.20 for cached input on the Kimi API, 33% cheaper than Kimi K3. The Highspeed version costs $4 / $20.
- Is Kimi K3.1 Pro open source?
- It is an open-weight model. Moonshot says the full weights will be released by Nov 4, 2026.
- How good is Kimi K3.1 Pro?
- Artificial Analysis ranks it 4th on its Intelligence Index at 55, behind only Claude Opus 5.5, Claude Sonnet 5.5 and Muse Wonder Full, and first on Terminal-Bench 4.0.
- How do I use Kimi K3.1 Pro?
- Pick it on kimi.ai or the Kimi app, in Kimi Work 3.1.0+, in Kimi Code with /model, or call kimi-k3.1-pro in the Kimi API.
Moonshot AI introduced Kimi K3.1 Pro on October 23, 2026, three months after Kimi K3. It is cheaper than K3 and much stronger on agent work: $2/$10 per million tokens, a 1M-token context window, and a Terminal-Bench 4.0 score within a point of Claude Opus 5.5.
Alongside it Moonshot shipped Kimi K3.1 Pro Highspeed, the same model at 2-3x the speed for twice the price. Both are live on Kimi.ai, the Kimi apps, Kimi Work, Kimi Code and the Kimi API today.
1. Kimi K3.1 Pro at a glance
Spec | Kimi K3.1 Pro |
|---|---|
Release date | Oct 23, 2026 (Oct 24 in China) |
Developer | Moonshot AI |
Parameters | 2.8T total, 104B active (MoE) |
Context window | 1,048,576 tokens |
Input | Text and image |
API price | $2 input / $10 output / $0.20 cached per 1M |
Highspeed price | $4 / $20 per 1M, 2-3x faster |
API model IDs | kimi-k3.1-pro, kimi-k3.1-pro-highspeed |
Open weights | By Nov 4, 2026 |
2. Kimi K3.1 Pro benchmarks
Moonshot's launch post leads with coding and agent tests. On Terminal-Bench 4.0, run with the Claude Code harness, K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and just behind Claude Opus 5.5 (66.4%). On Moonshot's own Kimi Code Bench 3.0 Long-horizon it scores 61.2%, up from K3's 55.95%.
Benchmark | Kimi K3.1 Pro | Comparison |
|---|---|---|
Terminal-Bench 4.0 (Moonshot) | 65.7% | Opus 5.5 66.4%, GPT-6 Astra 58.2% |
Terminal-Bench 4.0 (Artificial Analysis) | 67% | Sonnet 5.5 64%, Kimi K3 13% |
BenchCAD (with tools) | 92.7% | Opus 5.5 96.2%, GPT-6 Astra 95.9% |
OfficeQA Pro v2 | 64.4 | State of the art per Moonshot |
AA Intelligence Index | 55 (4th) | Opus 5.5 58, Sonnet 5.5 56 |
GDPval-AA | 1,846 Elo (3rd) | Opus 5.5 1,866 |
Humanity's Last Exam (AA) | 52% | Kimi K3 47% |
Hallucination rate (AA-Omniscience) | 44% | Kimi K3 53% |
Independent numbers from Artificial Analysis back up the agent story. Its Intelligence Index puts K3.1 Pro at 55, an 11-point jump over Kimi K3, and once the weights ship it becomes the highest-scoring open-weight model, narrowing the open-vs-closed gap from 12 points to 3. AutomationBench-AA, which tests SaaS workflows, rose from 58% to 73%.
Gains outside agents are smaller: Humanity's Last Exam moves from 47% to 52%, SciCode from 59% to 61%. Moonshot also cut hallucinations on its internal stress test from 23.4% to 8.3% at claim level, a change aimed at finance and research users.
3. Kimi K3.1 Pro price and API
On the Kimi API Platform, K3.1 Pro costs $2 per million input tokens, $10 per million output tokens and $0.20 per million cached input tokens, 33% lower than Kimi K3 on every rate. Highspeed doubles that to $4 / $20. Zero Data Retention is available for API customers.
Model | Input / 1M | Output / 1M |
|---|---|---|
Kimi K3.1 Pro | $2 | $10 |
Kimi K3.1 Pro Highspeed | $4 | $20 |
Kimi K3 | $3 | $15 |
Claude Sonnet 5.5 | $2 | $10 |
Claude Opus 5.5 | $4 | $20 |
Per task, the saving is smaller than the sticker suggests. Artificial Analysis measured $1.98 per Intelligence Index task, roughly the same as K3, because K3.1 Pro writes about 14% more output tokens (55K vs 48K). It matches Sonnet 5.5's list price and costs half of Opus 5.5.
The API model IDs are kimi-k3.1-pro and kimi-k3.1-pro-highspeed. The endpoint is OpenAI-compatible, and the model is also listed on OpenRouter as moonshotai/kimi-k3.1-pro.
4. How to use Kimi K3.1 Pro
- Chat: open kimi.ai or update the Kimi app on iOS, Android or HarmonyOS and choose K3.1 Pro.
- Desktop work: install Kimi Work 3.1.0 or later on Windows or Apple silicon Macs.
- Coding: run Kimi Code and switch with the /model command.
- API: create a key on platform.kimi.ai and call kimi-k3.1-pro or kimi-k3.1-pro-highspeed.
- Teams: Kimi Enterprise admins must switch K3.1 Pro on; it is off by default at launch.
- Self-hosting: wait for the open weights, due by Nov 4, 2026.
5. What Kimi K3.1 Pro is good at
Moonshot pitches four areas. Coding: full-project builds, game engines (Unity, Unreal, Godot) and polished Three.js front ends. 3D: parametric CAD and Blender scripts, plus a new 3D Make feature on kimi.ai. Finance: financial models, investment research and long business simulations, with 4.2x K3's score on its Business Operation test. Long-horizon agents: in one internal run it worked through 167 missions to build a walking-human simulator and fixed 36 physics errors on its own.
On cybersecurity, Moonshot says it removed cyber data from training before the open-weight release, so offensive capability stays below frontier closed models.
6. Where Sumus fits
Yes, you can put Kimi K3.1 Pro to work next to Claude, GPT and Gemini. In Sumus every model runs as an AI employee in your team chat, so you can send coding jobs to K3.1 Pro for its price, hand hard reasoning to Opus 5.5, and compare answers side by side without juggling keys and apps.
What we're watching
- The open-weight release, promised by Nov 4, 2026, and its license terms.
- Independent Terminal-Bench and SWE results beyond Moonshot's own harness.
- Whether Kimi K3.1 Flash or other K3.1 models follow.
FAQ
When was Kimi K3.1 Pro released?
October 23, 2026 (October 24 in China), on Kimi.ai, Kimi Work, Kimi Code and the Kimi API.
What is the Kimi K3.1 Pro context window?
1M tokens (1,048,576), with text and image input.
Kimi K3.1 Pro vs Kimi K3?
K3.1 Pro is 33% cheaper per token and 11 points higher on the Artificial Analysis Intelligence Index; Terminal-Bench 4.0 jumps from 13% to 67%.
Kimi K3.1 Pro vs Claude Opus 5.5?
Opus 5.5 still leads overall (58 vs 55 on the AA index) and edges it on Terminal-Bench 4.0, but K3.1 Pro costs half as much per token.
What is Kimi K3.1 Pro Highspeed?
The same model served 2-3x faster, at $4 / $20 per 1M tokens.
Is Kimi K3.1 Pro free?
You can try it in the Kimi app and on kimi.ai; API use is billed per token.
Sources
- Moonshot AI: Introducing Kimi K3.1 Pro, Oct 23, 2026
- Artificial Analysis: Kimi K3.1 Pro launch benchmarks, Oct 23, 2026
- OpenRouter model page moonshotai/kimi-k3.1-pro. Read Oct 28, 2026.
Update log
- — Oct 28: First published. Kimi K3.1 Pro released Oct 23, 2026; specs and price from Moonshot, independent scores from Artificial Analysis.