Moonshot AI API (Kimi models): pricing, free tier, rate limits and how to start

Use the Moonshot Kimi API with a current model identifier and a request interface that fits your application. Connect model choice, account controls and operational evidence before expanding a prototype.

What Moonshot offers through the API

Moonshot’s developer service is documented as the Kimi API Platform. Its current integration paths include OpenAI-compatible Chat Completions and Responses, plus an Anthropic-compatible Messages interface. Official documentation.

Treat Kimi API integration as a choice of both model and request contract. If your project already has a Chat Completions client, a small baseline can establish whether its required fields work with the Kimi route. Keep the response parser and optional features in that test. The same client class appearing in two examples does not establish that every parameter has the same behavior.

The present catalog positions Kimi K3 for coding and knowledge work, alongside the Kimi K2.7 Code family and the general-purpose Kimi K2.6. Its older Moonshot and Kimi identifiers are documented separately as retired. Official documentation.

That catalog history matters when copying an existing integration. A tutorial can look syntactically correct while selecting a retired identifier. Begin from a current model record and retain the model name in configuration. For an existing application, keep its passing fixtures before changing the identifier so the migration can be assessed against actual tasks rather than the appearance of a new release.

For knowledge work, build a fixture in which the answer can be checked against supplied material. Include missing facts and conflicting passages so the application’s behavior is visible when the request is difficult. For coding, require the patch to pass the project’s relevant checks. A provider’s positioning can justify adding a candidate to the evaluation; your acceptance test establishes whether it is suitable.

Kimi’s organization controls connect projects, member-created keys, shared account resources and project consumption settings. Official documentation.

Decide which project should own the experiment before making a credential. The administrative boundary becomes important when a prototype turns into a scheduled task or a shared coding service. Give that task a clear stop condition and document who can inspect its usage. A working local request is a useful first result, but it does not answer how an unattended worker should react when its account condition changes.

Use the sections below in the order that resolves your immediate uncertainty. Models explains identifiers and migration. Pricing turns the request shape into charge categories. The key guide establishes a controlled first call. Limits and Errors explain why a request might not complete even when the code is valid. The Python tutorial provides a compact baseline that you can extend deliberately.

Keep the first experiment narrow enough to inspect completely. Save its safe input, exact configuration, returned completion state and usage. Then change one meaningful element, such as the source material or output format, and compare the result. This creates a more durable integration record than a collection of successful console screenshots with no reproducible request behind them.

Pricing at a glance

Read Kimi API pricing for cache categories and workload planning.

Models available

Inspect current and retired Kimi models before reusing an older integration.

Free tier and how to get a key

Check recharge and voucher conditions and follow project key setup.

Rate limits and tiers

Plan shared organization capacity with Moonshot rate limits.

Service status

Start in 5 minutes

Follow the Kimi Python tutorial for the complete controlled example.

client = OpenAI(api_key=os.environ["MOONSHOT_API_KEY"], base_url="https://api.moonshot.ai/v1")
result = client.chat.completions.create(model="kimi-k3", messages=[{"role":"user","content":"Hello"}], reasoning_effort="low")

Recent changes

Use the AI API cost calculator to turn the model and workload you are considering into an estimate.

Last verified 12 Sep 2026 · Source ↗

Frequently asked questions

Sources

• Kimi API quickstart ↗
• Kimi model list ↗
• Inference pricing ↗
• Recharge and limits ↗
• Error reference ↗
• Organization management ↗
• Account and billing ↗
• Context caching ↗
• Batch API ↗
• Model parameters ↗

Last verified 12 Sep 2026 · Source ↗