Kimi API Token Pricing

ONE TOKEN

About 3 to 4 characters of English

CHEAPEST INPUT

$0.95 per million, Kimi K2.6 and K2.7 Code

OUTPUT

$4.00 to $15.00 per million, about 4 to 5× input

CACHED INPUT

80% to 90% cheaper, applied automatically

BATCH

40% off, Kimi K2.6 and K2.7 Code only

EXACT COUNT

Moonshot's Token Calculation API

From Moonshot's model pricing page and API reference, read on 8 October 2026.

A token is a small piece of text: a short word, part of a longer word, a space or a punctuation mark. Kimi reads and writes text as tokens, and Moonshot bills by counting them.

• Moonshot says 1 token is roughly 3 to 4 characters of typical English.
• So 1,000 English words are roughly 1,300 tokens, and a page of text about 650.
• Code, numbers, other languages and unusual words can use more tokens per character.

The split above is an illustration. Kimi's tokenizer may split the same text differently.

On a small screen, swipe the table sideways to see every column.

US dollars per million tokens, before tax. On Kimi K3 the 5-minute cache is used unless you choose 1 hour, and each cache hit refreshes the timer at no extra charge. From Moonshot's model pricing page.

On a small screen, swipe the table sideways to see every column.

Worked out from the per-million prices above. Moonshot does not say how it rounds very small charges.

On a small screen, swipe the table sideways to see every column.

Per million tokens, 60% of the standard price. Batch prices are listed only for these two models. From Moonshot's batch pricing.

Runs in your browser and does not send what you type to us. The box is marked to be hidden from the analytics tools this site uses.

A rough estimate using Moonshot's 3 to 4 characters per token for English; the middle figure uses 3.5. For other languages and code, use the exact counter below.

Every response includes a usage section that counts the tokens. Moonshot says cached, cache-written and uncached input add up to prompt_tokens.

"usage": {
  "prompt_tokens": 12000,
  "completion_tokens": 800,
  "total_tokens": 12800,
  "prompt_tokens_details": {
    "cached_tokens": 9000,
    "cache_write_tokens": 0
  }
}

Type the numbers from a response to see what it cost:

Standard prices before tax. Cache write is charged at Kimi K3's 5-minute rate.

Moonshot's Token Calculation API counts the tokens in a request before you send it, including images and video. It returns the total in data.total_tokens.

curl https://api.moonshot.ai/v1/tokenizers/estimate-token-count \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MOONSHOT_API_KEY" \
  -d '{
    "model": "kimi-k2.6",
    "messages": [{"role": "user", "content": "How many tokens is this?"}]
  }'

• Use it to check a long document fits the model's context window.
• Moonshot's page does not say whether this call is charged or rate limited.
• To cap the output, set max_completion_tokens. On Kimi K3 the default is 131,072. If the answer is cut off, finish_reason is "length".

From Moonshot's API reference. We have not run this request against the live API. Keys and setup: Kimi API setup.

1. 1

   Cap the output

   Output costs most. Set max_completion_tokens and ask for short answers.

2. 2

   Reuse the start

   Keep fixed instructions identical and first, so they are cached.

3. 3

   Send only what matters

   Trim old messages and send the relevant part of a document.

4. 4

   Think less on easy jobs

   Low reasoning effort on Kimi K3, thinking off on Kimi K2.6.