Kimi K3
Kimi K3 is an ultra-large-scale, open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.
Specifications
| Lab | Moonshot |
|---|---|
| Context window | 1M tokens |
| Input price | $3.00/1M |
| Output price | $15.00/1M |
| Release | Jul 2026 |
Strengths
- Supports text and image input with text output
- Million-token context window for very long tasks
- Tool-use support for agentic workflows
- Structured output support for reliable application integration
Weaknesses
- Newer listing with limited independent benchmark coverage
- Multimodal performance may vary by provider endpoint
Best for
- Agentic workflows and tool-using applications
- Multimodal document and screenshot analysis
- Long-context document and codebase processing
In Depth: Kimi K3
Summary
Kimi K3 is an AI model from Moonshot.
Released in Jul 2026. It takes text and image input and produces text output, with a context window of 1M tokens. Input pricing is $3.00 per 1M tokens and output is $15.00 per 1M tokens on OpenRouter.