What it does
The Kimi API Platform is the developer site for building with Kimi models. It lists three current models: K3, a flagship with a 1M-token context window aimed at software engineering, knowledge work and deep reasoning; K2.7 Code, a coding model with a 256k-token context window; and K2.6, a general-purpose model that accepts vision and text input, has thinking and non-thinking modes, and also has a 256k-token window. The platform offers official tools that plug into model calls, including web search, a Python code runner, a QuickJS JavaScript runner, Excel and CSV analysis, memory, URL fetch, date handling, unit conversion and Base64. It points to use cases in agent programming, deep research, legal document work and conversation analysis. Billing is pay-as-you-go, and rate limits rise automatically as cumulative spend grows. Enterprise options add flexible rate limits, SLA-backed reliability and dedicated support. The site also says Kimi K3 is available through several inference partners.
Capabilities
- API access to Kimi K3, K2.7 Code and K2.6
- 1M-token context on K3 and 256k-token context on the other models
- Official tools including web search, code runners and Excel analysis
- Pay-as-you-go billing with rate limits that grow with spend
- Developer console, documentation and playground
- Enterprise options with SLA-backed reliability
When to use it
- You are building coding or software-automation agents on a long-context model
- You need a model for deep research, legal document or conversation analysis workflows
- You want ready-made tools such as web search and a code runner available to the model
- You prefer usage-based billing for an API
