kimi-k3
kimimodel ID kimi-k3-42%42% OFF
kimi-k3 costs $1.74 per million input tokens and $8.70 per million output tokens through the KoRouter API — 42% below Kimi's official rate, pay as you go with credits.
Call it in one request
curl https://korouter.ai/v1/chat/completions \
-H "Authorization: Bearer sk-your-key" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k3",
"messages": [{ "role": "user", "content": "Hello" }]
}'kimi-k3 answers on the OpenAI-compatible endpoint, so anything that accepts an OPENAI_BASE_URL — Codex, Cline, Continue, Aider, Open WebUI, the OpenAI SDKs — needs the base URL, a key, and this model ID. Setup guides per tool.
Replace sk-your-key with a key from the console. Model IDs are bare names — see the docs.
What a real coding session costs
Per-token price tables mislead for agentic coding. A focused two-hour Claude Code session re-sends its context on every request, so 88% of its tokens are cache reads rather than fresh input — and the cache rate, not the headline input price, decides most of the bill. Priced at kimi-k3's live rates, that session costs $2.78 through KoRouter, against $4.80 at Kimi's official prices.
- Cache reads · 5M tokens$0.87
- Fresh input · 250K tokens$0.43
- Output · 100K tokens$0.87
- Cache writes · 350K tokens$0.61
- Session total$2.78
The token mix is a stated assumption, not a measurement — change it and re-price every model in the cost calculator, or read the full worked example in what a real Claude Code session costs. Prices come from the live billing catalog.
The same session elsewhere
The same reference session, priced at each model's live KoRouter rates — the nearest options on either side of kimi-k3.
- gpt-5.6-sol$1.35
- claude-sonnet-5-5$1.87
- claude-sonnet-5$2.19
- claude-opus-5-5$3.74
- claude-opus-4-6$5.48
- claude-opus-4-7$5.48
More models
Prices use the live billing catalog · live availability · all models