Kimi K3 API

Get started with the hosted Kimi K3 API — access, pricing, and a minimal example.

Access

Select kimi-k3 on the Kimi API Platform. Context caching is automatic, with reported cache hit rates above 90% on coding workloads, lowering effective input cost.

Pricing

Cache-hit input	$0.30 / 1M tokens
Cache-miss input	$3.00 / 1M tokens
Output	$15.00 / 1M tokens

Minimal request shape

A standard chat-completions style call, specifying the kimi-k3 model and your messages. Always verify the exact request format against the official Kimi API documentation before production use.

{
  "model": "kimi-k3",
  "messages": [
    { "role": "user", "content": "Summarize this PDF and draft a team update." }
  ]
}

Related: pricing, token calculator.

KimiK3 Max is an independent fan and reference site. It is not affiliated with, endorsed by, or operated by Moonshot AI. “Kimi” is a trademark of Moonshot AI. All trademarks, model names, and benchmark data belong to their respective owners.

Frequently Asked Questions

How do I access the Kimi K3 API?

Sign in to the Kimi API Platform and select the kimi-k3 model. Pricing is $0.30 / $3.00 / $15.00 per million cache-hit input / cache-miss input / output tokens.

Does Kimi K3 support context caching?

Yes. Caching is automatic (no cache ID required). Moonshot reports cache hit rates above 90% on coding workloads, which sharply lowers effective input cost.

Is the Kimi K3 API open weight?

The API is a hosted service. The open-weight release (Modified MIT, expected July 27, 2026) is separate and intended for self-hosting.