Project scaffold
Build
Full Project Coding
Turn a brief into an architecture plan, file map, and starter implementation.
Kimi K3.1 Pro is liveThe independent workbench
The independent workbench
for Kimi K3.1 Pro
Build, review, and analyze with a model optimized for long-context coding and multimodal reasoning. Upload screenshots, videos, documents, or entire project briefs and get structured, exportable results.
Unofficial independent platform. Not affiliated with Moonshot AI.
Near-Opus coding
Generate and review full project code with a model built for long-horizon software work.
1M-token context
Load large documents, repository briefs, and research sets without losing the thread.
Image & video input
Upload screenshots, UI mockups, product videos, or walkthroughs for structured analysis.
Open weights soon
Prepare for local deployment when Kimi K3.1 Pro weights become available.
ChatFilesArtifacts
Ask anything or drop files…
Workbench scenarios
Built for project-level work, not casual chat
Every mode is designed around a concrete developer task, with structured output and reusable history.
Code review
Review
Codebase Audit
Ask for risks, missing tests, dependency issues, and refactor priorities.
Document analysis
Analyze
Long Document Research
Extract requirements, compare contracts, or synthesize dense technical material.
Multimodal input
IMG
VID
Understand
Image & Video Analysis
Convert UI screenshots into frontend specs and product videos into action items.
Agent plan
Goal
Research
Code
Verify
Plan
Agent Task Planning
Break complex goals into tool-ready steps with success criteria.
3D scene graph
Create
3D & Game Systems
Prototype scene structures, shaders, game loops, and interaction systems.
Open weights countdown
Local deployment is next
When the weights open, K3.1 Studio will support connecting your own endpoint and running the same workbench locally.
00
days
00
hours
00
minutes
00
seconds
Self-host roadmap
Bring your own infrastructure
Phase 1
API workbench
Use the hosted service with transparent credit metering.
Phase 2
Own endpoint
Point K3.1 Studio to your own Kimi-compatible inference server.
Phase 3
Local weights
Run fully on your hardware with local history and artifacts.
Comparison
Where K3.1 Pro fits
A concise, non-marketing view of what matters for long-context and multimodal development.
| Capability | Kimi K3.1 Pro | Claude Opus | GPT-6 | Gemini |
|---|---|---|---|---|
| Long-context coding | 1M tokens | Large | Large | Very large |
| Image & video understanding | Yes | Image-only | Image & video | Image & video |
| Open weights | Coming Nov 4 | No | No | Partial |
| Independent workbench | Yes | — | — | — |
Prompt templates
Start with a task, not a blank box
Template
Project scaffold
Generate architecture, file tree, milestones, and starter code from a one-paragraph brief.
Template
Screenshot to spec
Convert a UI image into component structure, states, and accessibility notes.
Template
Video walkthrough
Extract product flows, bugs, and improvement ideas from a recorded demo.
FAQ
Questions, answered
Is this official?
No. K3.1 Studio is an unofficial independent platform and is not affiliated with Moonshot AI.
What can I upload?
Images, short videos, and documents. Larger files may be limited based on your plan and API policy.
How are credits used?
Credits are based on message length and attachment type. You always see the estimate before sending.
Can I use my own endpoint later?
Yes. The roadmap includes connecting your own Kimi-compatible endpoint when open weights are available.