None
None
Long Story Video Skill
Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.
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
Feedback
More AI Tools
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
AI Video Effects
Loading...
All Tools
Browse AI tools
100% Free Image Generator
Free GPT Image 2.5
Seedance 2.5
MiniMax H3 video generator
Seedance 2.0 AI Video Generator — The Future of AI Video Is Here
AI Video Effects
AI Effects - Create Funny Videos Easy!
Kimi K3.1
Kimi K3.1 is an open-weight 3T model with a one-million-token window, post-trained for programming, 3D modelling and financial research. It writes and runs Blender scripts, ships whole projects, and keeps the reasoning behind every number.
What Is Kimi K3.1?
Kimi K3.1 is the post-trained successor to Kimi K3: a 3-trillion-parameter open-weight model with a 1M-token context window. Far more reinforcement-learning compute went into training across hundreds of downstream domains, so the gains show up in real work rather than in a single chart. It builds whole projects instead of snippets — front-end pages, Three.js scenes, Unity and Godot prototypes — and in 3D it writes and runs Blender and CAD scripts to produce parametric parts, edit geometry, and lay out scenes. On financial work it scores 64.4 on OfficeQA Pro v2, the strongest result recorded, and claim-level hallucination on the internal hard set falls from 23.4% to 8.3%.
- 3T Open Weights, 1M ContextKimi K3.1 carries a million-token window, enough to hold an entire repository or a filing bundle, and the weights are being released for teams that want to run it themselves.
- Long-Horizon ProgrammingSmall fixes in a live codebase or a project built from nothing: Kimi K3.1 scores 65.7% on Terminal-Bench 4.0 and 61.2% on Kimi Code Bench long-horizon, ahead of the K3 it replaces.
- Parametric 3D, No ConnectorIt writes and runs Blender and CAD scripts directly, so mechanical parts, furniture and architectural models come back as editable parametric geometry instead of a picture of one.
How to Use Kimi K3.1
Three steps: describe the job, attach the files or repository it should read, then let it plan, execute, and check its own work. Kimi K3.1 runs tools inside the conversation, so a prototype, a 3D asset or a model workbook comes back as a finished artefact rather than a to-do list.
Why Builders Choose Kimi K3.1
Most assistants stop at a plausible answer. Kimi K3.1 keeps going until something runs: it edits the code, opens the 3D viewer, re-reads the filing, and reports where the numbers came from. Reinforcement learning across hundreds of domains is what makes that possible, and the effect is measurable — 65.7% on Terminal-Bench 4.0, 92.7% on BenchCAD, and a claim-level hallucination rate on the internal hard set cut from 23.4% to 8.3%.
A Million Tokens in View
The 1M-token window holds a whole repository, a contract set or a quarter of filings. Kimi K3.1 keeps the detail it read on page one available when it answers on page nine hundred.
Code That Runs
Long-horizon coding is the headline: 65.7% on Terminal-Bench 4.0, a walk through real terminal tasks under resource limits, plus game-engine and Three.js work that comes back playable.
3D Straight from the Chat
Sketch, brief or reference image in; parametric model out. Kimi K3.1 scripts Blender and CAD itself, edits components, and exports printable files, video or a Live Photo.
Finance You Can Audit
OfficeQA Pro v2 puts it at 64.4, the best score recorded, with modelling up 7.2 points and research up 11.6 over K3. Every figure traces back to a source document.
Two Speeds, One API
Pick kimi-k3.1-pro for depth or kimi-k3.1-pro-highspeed for two to three times the pace, with zero data retention for API customers and enterprise accounts kept isolated by default.
Premium Plans & Pricing
Choose the perfect plan for your AI Video/Image Generator needs and start creating professional videos.
Basic-Yearly
$0.015/credit
Includes these features:
- 7200 credits per year
- High quality AI image Generator
- Powerful AI Video Generation
- AI video effect features
- Commercial License
- Fast generation queue
- Private generations
- Valid for 1 year
- Cancel anytime
/Year
Pro-Yearly
$0.013/credit
Includes these features:
- 25200 credits per month
- High quality AI image Generator
- Powerful AI Video Generation
- AI video effect features
- Commercial License
- Fast generation queue
- Private generations
- Valid for 1 year
- Cancel anytime
/Year
Max-Yearly
$0.01/credit
Includes these features:
- 54000 credits per year
- High quality AI image Generator
- Powerful AI Video Generation
- AI video effect features
- Commercial License
- Fast generation queue
- Private generations
- Valid for 1 year
- Cancel anytime
/Year
Kimi K3.1 — Frequently Asked Questions
Answers on what Kimi K3.1 is, how long its context runs, how it scores on coding and 3D benchmarks, whether the weights are open, and where to use it today.
What is Kimi K3.1?
Kimi K3.1 is the first model in the K3.1 line: a 3-trillion-parameter open-weight model with a 1M-token context window, post-trained for programming, 3D work, financial research and agentic tasks.
How large is the context window?
One million tokens. That is enough to keep an entire codebase, a long transcript or a set of filings in view for the whole conversation instead of chunking it away.
How does Kimi K3.1 score on coding?
65.7% on Terminal-Bench 4.0, against 58.2% for GPT-6 Astra, and 61.2% on Kimi Code Bench long-horizon, up from 55.95% for K3. Benchmarks run at max reasoning effort with temperature 1.0.
Can Kimi K3.1 build 3D models?
Yes. It scores 92.7% on BenchCAD, writing Blender and CAD scripts without a connector. The 3D Make feature on kimi.ai previews and edits the result, then exports printable files, video or Live Photo.
Are the weights open?
Kimi K3.1 ships as an open-weight release, with full weights due by November 4, 2026. Cybersecurity data was removed from both training stages, and the team states the model trails frontier closed models on offensive security.
Where can I use Kimi K3.1?
Through Kimi Agents on the web and mobile apps, Kimi Work on desktop, Kimi Code via /model, and the API under kimi-k3.1-pro or kimi-k3.1-pro-highspeed. Business accounts are isolated and off by default.
Put Kimi K3.1 on Your Next Build
Hand over the repository, the sketch or the filings and let Kimi K3.1 carry the work to something runnable. Long context, real tools, and a million tokens of source material still in view when the answer lands.