Moonshot
Kimi K3
Kimi K3 offers three reasoning levels — Low, High, and Max — so you can trade response speed for thinking depth.
Quick facts
Pricing
Input $3 / Output $15 per 1M tokens
Context window
1.05M
Max output
1.05M
Knowledge cutoff
Unknown
Reasoning levels
Kimi K3 comes with 3 selectable reasoning levels. Pick the level right in the crowSync composer — higher levels think longer on hard problems (and use more output tokens); token prices stay the same at every level.
Low
Kimi K3 Low
High
Kimi K3 High
Max
Kimi K3 Max
What this model can do
You provide
- 📄 Documents (PDFs, Word, TXT)
- 🖼️ Images & Screenshots
- 💬 Text instructions
It creates
- 📝 Formatted text & summaries
- 📊 Excel sheets & CSV data
- 📽️ PowerPoint slides & decks
- 💻 Usable code snippets
- 🤖 JSON & structured data
- 🎨 Generated images
Capabilities
Reasoning
Better suited to multi-step analysis, planning, and harder prompts.
Vision
Can work with screenshots, charts, or other image inputs.
Document input
Can accept supported documents directly in chat.
Structured output
Can return JSON for workflows that need machine-readable output.
Tool use
Can use compatible tools when crowSync enables them for the run.
Native tools in crowSync
Web search
Search the live web during a run.
Available
Web fetch
Open and read webpage contents directly.
Available
Code interpreter
Run code for calculations, analysis, and generated files.
Available
Image generation
Generate images directly from prompts.
Available
Input and output
Input
- Text
- Image
- Documents
Output
- Text
Common use cases
- Complex planning, debugging, and multi-step analysis.
- Research tasks that need current information from the web.
- Summarizing PDFs, documents, or longer working files.
- Data analysis, calculations, and code-backed workflows.
Pricing details
Input
$3 / 1M tokens
Output
$15 / 1M tokens
Cached input
$0.30 / 1M tokens