Kimi K3 is now available through the Amanai API, dashboard calculator, documentation, status page, and live CLI model pickers.
What changed
A new public model ID, amanai/kimi-k3, is available anywhere Amanai models can be selected. Existing API keys and OpenAI-compatible request formats continue to work without configuration changes.
Capabilities
Kimi K3 supports long-context text generation, configurable reasoning effort, streaming responses, and function tool calls. It is available through the same Chat Completions and Responses interfaces used by the rest of the Amanai text catalog.
Kimi K3 supports up to 1M tokens of context. Actual usable context depends on request structure, output allowance, and upstream availability.
Using Kimi K3
Set the request model to amanai/kimi-k3:
curl https://api.amanai.dev/v1/chat/completions -H "Authorization: Bearer $AMANAI_API_KEY" -H "Content-Type: application/json" -d '{"model": "amanai/kimi-k3","messages":[{"role": "user","content": "Hello"}]}'Billing
Kimi K3 uses Pricing Engine v2 with separate rates for uncached input, generated output, and cached input. Its current multipliers are:
| Component | Multiplier |
|---|---|
| Input | 6.47 |
| Output | 32.34 |
| Cached input | 0.32 |
The rate is positioned slightly above GPT 5.6 Sol. This is a current catalog addition and does not retroactively alter prior usage or rewrite the historical Pricing Engine v2 launch.
Availability
Kimi K3 appears automatically in the model API, dashboard calculator, documentation, status page, Codex setup, and Claude Code model picker. No new API key is required.