The Kimi API has no monthly fee. You top up a balance and pay only for the text that goes in and the text that comes out, measured in tokens. This page lists the price of every model, shows what real jobs cost, and gives you a calculator for your own numbers.
Bars drawn to scale. Kimi K3 costs about three to four times as much as the other two.
Kimi K2.6 and K2.7 Code: $0.95 in and $4.00 out per million tokens.
Kimi K3: $3.00 in and $15.00 out per million tokens.
Top up from $1. No subscription. Pay only for what you use.
A token is a small piece of text, roughly 3 to 4 characters of English. A full page is about 650 tokens.
Input is what you send. Output is what the model writes back. Output costs about four to five times more.
You add money to a balance and each request draws from it. Kimi memberships are a separate product.
New to tokens? See how Kimi works.
| Model | Input | Cached input | Output | Reads at once |
|---|---|---|---|---|
| kimi-k3The most capable | $3.00 | $0.30 | $15.00 | 1,048,576 |
| kimi-k2.7-codeMade for coding | $0.95 | $0.19 | $4.00 | 262,144 |
| kimi-k2.7-code-highspeedThe same, faster | $1.90 | $0.38 | $8.00 | 262,144 |
| kimi-k2.6Everyday general model | $0.95 | $0.16 | $4.00 | 262,144 |
On a small screen, swipe the table sideways to see every column.
US dollars per million tokens, before tax. "Cached input" is text the model has seen before. "Reads at once" is the context window, in tokens. From Moonshot's model pricing page.
Guides for each: Kimi K3 API · K2.7 Code API · Kimi K2.6 API
Prices per million tokens are hard to picture. Here are three everyday jobs worked out at list prices.
A short question and answer on Kimi K2.6.
1,000 tokens in × $0.95
300 tokens out × $4.00
One long document read by Kimi K3.
65,000 tokens in × $3.00
2,000 tokens out × $15.00
A busy session on Kimi K2.7 Code.
2 million tokens in, 80% cached
100,000 tokens out
The token counts are our own assumptions, chosen to illustrate. They are not measurements, and your usage will differ.
Enter a month's usage. Token amounts are in millions.
List prices before tax. It leaves out Kimi K3's one-off charge for saving text to the cache and any web search calls.
| Batch price | Input | Cached input | Output |
|---|---|---|---|
| kimi-k2.7-code | $0.57 | $0.114 | $2.40 |
| kimi-k2.6 | $0.57 | $0.10 | $2.40 |
On a small screen, swipe the table sideways to see every column.
Search tools are charged per call, on top of any tokens the model uses.
| Tool | Price per call | What it does |
|---|---|---|
| Web Search Basic | $0.002 | Searches the web and returns results |
| Web Search Pro | $0.003 | Adds site and date filters and tidier results |
| URL Fetch | $0.002 | Reads one web page and returns its text |
On a small screen, swipe the table sideways to see every column.
$web_search, costs $0.005 per call. Moonshot expects to retire it on 20 October 2026 and recommends the newer tools.More: Web Search API pricing.
When your total top-ups reach $5, Moonshot adds a $5 voucher.
Pulling text out of files and storing them is temporarily free. You still pay for the tokens when that text goes to a model.
Search and fetch calls that fail or return nothing are not charged.
There is no free tier for the models themselves. To use Kimi without paying, use the free plan in the app: Is Kimi AI free?
How much you have topped up in total decides how fast you may send requests. The minimum top-up is $1.
| Total top-ups | At once | Requests a minute | Tokens a minute |
|---|---|---|---|
| $1 | 1 | 3 | 500,000 |
| $10 | 15 | 100 | 2,000,000 |
| $20 | 40 | 100 | 3,000,000 |
| $100 | 50 | 200 | 3,000,000 |
| $1,000 | 60 | 200 | 4,000,000 |
| $3,000 | 100 | 300 | 5,000,000 |
On a small screen, swipe the table sideways to see every column.
Need more than the top tier, a contract or a volume discount? See enterprise Kimi API pricing.
Use Kimi K2.6 for routine work and keep Kimi K3 for the hard cases.
Keep the fixed part identical and first, so it is charged at the cached rate.
Overnight jobs on Kimi K2.6 or K2.7 Code cost 40% less.
Set an output limit on every request. Output is the expensive part.
On Kimi K3, set the reasoning effort to low for simple jobs. On Kimi K2.6, switch thinking off.
Trim old messages and send only the parts of a document you need.
The two are billed separately, and a membership does not include API credit. More: API vs subscription · Kimi AI pricing
We could not find these details on the pages we read. Check them on Moonshot's platform or ask its support before you rely on them.