Kimi K3.1 Pro Workspace
Describe what you need — a script, a shader, a CAD part or a whole landing page — and Kimi K3.1 Pro drafts it; switch to Highspeed whenever the clock is tight.
None

None

None

Long Story Video Skill

Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.

Ads Video Skill

Ads Video Skill

Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.

3D Science Explainer Video Skill

Convert scientific concepts into stunning 3D explain animations

AI Video Prompt Generator

Feedback

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

kimi k3.1

Kimi K3.1 answers faster, remembers longer and builds more: trillion-parameter open weights, a million-token memory and a Highspeed mode for quick turns.

All Tools

Discover our comprehensive AI-powered animation toolkit

The Edge Behind Kimi K3.1 Pro

The K3.1 family opens with Kimi K3.1 Pro — one million tokens of working memory, weights released openly at 3T scale, and measurable gains in code, 3D and finance research.

  • A trillion-parameter release you can inspect
    Kimi K3.1 Pro tops the K3.1 family: the largest open-weight model Kimi has shipped at 3T scale, holding roughly a million tokens inside a single session.
  • Highspeed variant, 2-3x quicker replies
    A sibling build — Kimi K3.1 Pro Highspeed — returns answers two to three times faster than the standard checkpoint, which keeps rapid back-and-forth workable.
  • Joint reinforcement learning across hundreds of fields
    Trained on from K3 with much deeper post-training, Kimi K3.1 Pro applies joint RL over hundreds of downstream domains, so every one of them improves together.

How to Run Kimi K3.1 Pro

From a blank prompt to a shipped build in three moves with Kimi K3.1 Pro.

What Kimi K3.1 Pro Can Do

Coding agents, CAD and 3D authoring, finance pipelines and measured cyber safeguards — the areas where Kimi K3.1 Pro pays off.

Coding agents that run for hours

Terminal-Bench 4.0 scores Kimi K3.1 Pro at 65.7% — ahead of GPT-6 Astra (58.2%) and a hair behind Claude Opus 5.5 (66.4%) — while Kimi Code Bench 3.0 Long-horizon rises to 61.2% from K3's 55.95%.

Game engines and Three.js scenes

Unity, Unreal Engine and Godot are all supported alongside web search and 3D generation, and its front-end output shows stronger layout, typography and motion.

Parametric 3D and CAD scripting

BenchCAD with tools reaches 92.7%; it also writes Blender or CAD scripts to build assets, swap parts and tune geometry, materials and layout — with no MCP connectors required.

Finance work from start to finish

OfficeQA Pro v2 lands at 64.4; measured against K3 it gains 7.2 points on Financial Modeling, 11.6 on Financial Research and about 4.2x on Business Operation.

Accuracy that holds under pressure

On the internal Kimi Hallu Bench Hard, claim-level hallucination falls from 23.4% on K3 to 8.3%, and replies containing at least one factual slip drop to 42.4% of K3's rate.

Cyber ability kept on a leash

Cyber material was removed from mid-training and RL; the shipped model still clears 81.0% of xdev-bench-hard at pass@3, against 90.5% for the unfiltered checkpoint.

FAQ

Kimi K3.1 Pro: Common Questions

Answers on Kimi K3.1 Pro — what it is, how the Highspeed build differs, when the weights land and where you can run it today.

1

So what is Kimi K3.1 Pro, exactly?

The opening release in Kimi's K3.1 series: K3 taken through far heavier post-training, then shipped as openly available weights at 3T scale with a million-token context.

2

How far ahead of Kimi K3 is it?

The jump shows up across coding, agent and knowledge benchmarks — 61.2% against 55.95% on Kimi Code Bench 3.0 Long-horizon, with hallucination down from 23.4% to 8.3%.

3

When will the weights be downloadable?

The full checkpoint is scheduled for release by Nov 4, 2026; hosted access through Kimi's own products already works today.

4

Where can I try it right now?

On kimi.ai and the mobile apps, in Kimi Work 3.1.0 or later, in Kimi Code, and via the Kimi API, which offers Zero Data Retention.

5

Is there something faster?

Yes — the Highspeed build replies two to three times quicker and is callable as kimi-k3.1-pro-highspeed on the Kimi API Platform.

6

How does it stack up against closed frontier models?

It trades blows with the best proprietary systems: 65.7% on Terminal-Bench 4.0 (Claude Opus 5.5 at 66.4%, GPT-6 Astra at 58.2%) and 92.7% on BenchCAD with tools.

Start Building with Kimi K3.1 Pro

Run Kimi K3.1 on Kimi.ai or through the API, reach for Highspeed when latency matters, and watch for the full weights by Nov 4, 2026.