Input
- Official Moonshot
- $3
- Here (−50%)
- $1.5
k3
Kimi K3 API — price per token
Kimi K3 is Moonshot's frontier model — a 1M-token window for agentic coding and long-context work, addressed as k3 on the Anthropic Messages lane.
Pricing per 1M tokens
| Rate | Official Moonshot | Here (−50%) |
|---|---|---|
| Input | $3 | $1.5 |
| Cached input | $0.3 | $0.15 |
| Cache write | $3 | $1.5 |
| Output | $15 | $7.5 |
Cached input
- Official Moonshot
- $0.3
- Here (−50%)
- $0.15
Cache write
- Official Moonshot
- $3
- Here (−50%)
- $1.5
Output
- Official Moonshot
- $15
- Here (−50%)
- $7.5
Every request is metered at the official rate first, then the flat 50% B2C discount is subtracted before it touches your prepaid balance. Context window: 1M tokens. Max output: not published.
Best for
- Agentic coding sessions that outgrow a 256K window.
- Long-context analysis across a whole repository or corpus.
- Work already configured for Claude Code, which selects this window as k3[1m].
Good to know
- Cached input bills at 10% of input — caching is automatic on repeated prefixes.
- KIMI publishes no separate cache-write rate: a write is a cache miss and bills at the input rate.
How to use Kimi K3
Create a free account, generate one key, and point any Anthropic-compatible tool at https://router.apitoken.sale with model ID kimi/k3, authenticated with x-api-key. Kimi speaks the Anthropic Messages protocol, so the catalogue namespace in the model ID is what selects it — the base URL is the same one Claude uses. Eligible new accounts include $5 of platform bonus credit — enough to test the model before topping up.
Frequently asked questions
How much does the Kimi K3 API cost?
Officially $3.00 per 1M input tokens and $15.00 per 1M output tokens, with cached input at $0.30. On apiToken.sale the same requests cost 50% less — $1.50/$7.50 at the flat discount applied to every call.
What is the model ID for Kimi K3?
k3, or kimi/k3 on the unified router. Claude Code users can also send k3[1m], which is that client's spelling for the same 1M window — both settle identically.
Is there a 256K variant?
Yes: k3-256k is the same model and the same rate card with a 256K accepted window. Choose it when a smaller window is what your harness expects.
Run Kimi K3 on the Anthropic Messages API at a flat 50% off — instant key, prepaid balance, card or crypto.