API Cost Check
Kimi API cost calculator
Moonshot prices Kimi per token with input, output and cached-input rates; a cache hit is billed at 10%–20% of the input price in our data, with no write fee listed. Moonshot also offers an explicit context cache paid by storage time, which is not in our data; the tables use the per-token cached-input price only.
No batch price is listed for Kimi in our data, so the Batch column repeats the standard price. The spread between models is wide: Kimi K3 costs 5× as much as Kimi K2.5 for the same traffic. Use the flagship only where its longer context or harder reasoning pays off, and the smaller models for chat and routine tasks.
The table prices one workload on each model: 10,000 requests a day with 6,000 input and 500 output tokens per request. Columns: no caching; 70% of input read from the cache and 5% written to it; every request sent through the Batch API. For the same traffic Kimi K3 costs 5× as much as Kimi K2.5.
| Model | Standard 10,000 req/day · 6,000 in · 500 out Calculate with your numbers → | With prompt caching 10,000 req/day · 6,000 in · 500 out · cache 70% Calculate with your numbers → | Batch API 10,000 req/day · 6,000 in · 500 out · batch Calculate with your numbers → |
|---|---|---|---|
| Kimi K3 | $7,650/mo | $4,248/mo | $7,650/mo no batch |
| Kimi K2.7 Code | $2,310/mo | $1,352/mo | $2,310/mo no batch |
| Kimi K2.6 | $2,310/mo | $1,315/mo | $2,310/mo no batch |
| Kimi K2.5 | $1,530/mo | $900/mo | $1,530/mo no batch |
Estimates from published per-token rates; check official pricing before budgeting. Data synced from the LiteLLM price list 2026-10-12 08:00 UTC.
Cost per 1,000 requests
For 1,000 requests of 6,000 input and 500 output tokens, Kimi K2.5 costs $5.10 at standard rates, $3.00 with caching and $5.10 through the Batch API; Kimi K3 costs $25.50, $14.16 and $25.50. Multiply by your own request count, or put your traffic into the calculator.
How to cut the bill, in order
- Keep the shared prefix at the start of every prompt so cached-input pricing applies; only the user turn should change.
- Use the smaller models for chat and routine tasks; reserve the flagship for long documents and hard reasoning.
- Prefer retrieval over sending whole documents unless the same document is cached and queried many times.
- Cap output length; reasoning tokens are billed as output.
- Count prompts with the token counter, check your hit rate in the caching calculator, then budget in the calculator.
When does caching pay off for your traffic? Use the prompt caching calculator →
FAQ
Does Kimi charge for cache writes?
No per-token write fee in our data: a miss is billed as normal input and hits at the cached-input price. The explicit cache has a time-based storage charge that is not in our data.
Is a 1M-token context worth paying for?
Only when one request really needs the whole document and the same document is asked about many times with caching. Otherwise retrieval with a few chunks costs far less per question; see the RAG cost guide.
Does Kimi charge for reasoning tokens?
Yes, as output tokens. Measure the output token count of real responses and use that, not the visible answer length, in the calculator.
Which Kimi model for coding?
Compare the code model with the flagship on cost per completed task: agents resend a growing context, so cached-input price and context window matter more than the per-token headline. Use the agent cost calculator with your calls per task.