Skip to content

Kimi API cost calculator

Enter your token volume and cache hit rate to see what each Kimi model would actually cost. Results update as you type, run entirely in your browser, and use pricing taken directly from Moonshot's documentation.

Last updated

Your workload

Share of input tokens served from the prompt cache. Agents with a fixed system prompt typically sit between 60% and 90%.

Cheapest option

Kimi K2.5

$130/mo

Not available to new accounts — check availability before planning around this price.

Most expensive

Kimi K3

$630/mo

Sorted cheapest first. Pricing from platform.kimi.ai, verified 6 August 2026. Excludes tax.
ModelStatusCost per requestMonthly costvs cheapest
Kimi K2.5Closing 2026-08$0.0130$130baseline
Kimi K2.6Available$0.0191$1911.5×
Kimi K2.7 CodeAvailable$0.0194$1941.5×
Kimi K2.7 Code HighSpeedAvailable$0.0388$3883.0×
Kimi K3Available$0.0630$6304.8×

Showing estimated costs for 10,000 requests per month.

FAQ

Working out Kimi API costs

How is Kimi API cost calculated?

Cost is the sum of input and output tokens, each billed per million. Input is split into cache hits and cache misses at different rates. The formula is: (input tokens × cache hit share × hit rate + input tokens × miss share × miss rate + output tokens × output rate) ÷ 1,000,000.

What is a cache hit and why does it matter?

Moonshot automatically caches repeated prompt prefixes. When a request reuses a prefix the model has already seen, those input tokens bill at the cache-hit rate — as little as one tenth of the miss rate. For agents that resend the same system prompt and tool definitions every call, this is the single largest lever on your bill.

What cache hit rate should I assume?

For a chatbot with varied prompts, assume 0 to 20 percent. For an agent with a fixed system prompt and tool schema, 60 to 90 percent is realistic because the prefix repeats on every call. When in doubt, model both ends and treat the answer as a range.

How many tokens is my prompt?

For typical English text, one token is roughly 3 to 4 characters, so 1,000 words is about 1,300 to 1,500 tokens. Code and non-English text tokenize less efficiently. Use our token counter for an estimate, or Moonshot’s token estimation endpoint for an exact count.

Do these prices include tax?

No. Moonshot publishes prices excluding tax, and your actual tax depends on your billing jurisdiction. Treat every figure here as a pre-tax estimate.