How to Track Kimi / Moonshot AI: Models, API, Kimi Code, and Pricing
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
As of July 2, 2026, Kimi/Moonshot should be evaluated as four separate builder surfaces: long-context Chinese material work, API integration, pricing and limits, and Kimi Code as a coding-agent workflow.
Bottom line
| Decision | Check today | Good pilot signal | Stop signal |
|---|---|---|---|
| Long Chinese documents | Kimi model/API docs and file workflow | Output keeps sources, caveats, and next actions separate | Fluent summary without traceable evidence |
| API integration | Overview, model list, chat completion, tool use, limits | Fits existing gateway and error logging | Only a demo works; monitoring is unclear |
| Cost check | Pricing, recharge, limits, batch/tool cost | Cost is calculated per task with retries and review time | Only token price is compared |
| Kimi Code | GitHub README, getting started, ACP/IDE docs | Small repo diff, commands, tests, and rollback notes | Broad edits, unclear permission, no rollback |
Use cases
Use Kimi for a small long-document pilot, a fixed API classification task, and a low-risk Kimi Code repo trial before making a broader adoption call.
Not good for
Do not use this page as a permanent price table or a single winner claim. Recheck official docs before quoting model availability, cost, or install steps.
Example
A PM can test a Chinese report summary with required source notes; an engineer can test a 20-item JSON classification request; a developer can ask Kimi Code to explain a small repo before editing README or tests.