Kimi K2.7 Code is Moonshot AI's model made for programming. You give it a coding task, and it reads your files, makes changes, runs tests and keeps going until the job is done. It is the model behind much of Kimi's coding work today. This page explains what it does, what it costs and how to use it.
An illustration of the working loop, not a recording of a real session.
Kimi's dedicated coding model, built on Kimi K2.6 and tuned for long tasks.
$0.95 in and $4 out per million tokens on the API. HighSpeed is double.
Through the Kimi Code tool, the API, or by running the open weights yourself.
Most AI models are generalists. Kimi K2.7 Code is a specialist. Moonshot took its general model, Kimi K2.6, and trained it further for one kind of work: software tasks that take many steps. Moonshot describes it as following instructions more reliably in long sessions and finishing coding tasks more often.
From Moonshot's model card, models page and launch announcement. Under the hood it is the same size as Kimi K2.6: about 1 trillion parameters, with 32 billion active at a time.
Moonshot says it succeeds more often on very long runs: sequences of more than 4,000 tool calls and more than 12 hours of continuous work.
It uses about 30% fewer thinking tokens than Kimi K2.6. That means lower bills and faster answers for the same work.
Thinking is switched on permanently. There is no instant mode, unlike Kimi K2.6.
HighSpeed runs the same model at about 180 tokens a second, for people who want quicker responses and will pay more.
Moonshot published six test results comparing Kimi K2.7 Code with Kimi K2.6. Higher is better. It improved on all six.
Moonshot's own coding test
Writing whole programs
Machine-learning engineering tasks
Moonshot's own long-running agent test
Using outside tools
Using outside tools, verified
For outside views, see our Kimi AI review and the comparison hub.
| Standard | HighSpeed | |
|---|---|---|
| API name | kimi-k2.7-code | kimi-k2.7-code-highspeed |
| Speed | Normal | About 180 tokens a second, up to 260 in short conversations |
| Input price | $0.95 | $1.90 |
| Repeated input | $0.19 | $0.38 |
| Output price | $4.00 | $8.00 |
| Quality | The same model, so the same answers | |
| Pick it when | Cost matters, or jobs run in the background | You are waiting at the keyboard and time matters |
On a small screen, swipe the table sideways to see every column.
Prices are per million tokens, before tax, from Moonshot's price page. More: K2.7 Code HighSpeed.
Moonshot's coding tool for your terminal. The route Moonshot recommends. Needs a paid Kimi membership.
Call the model from your own software or another coding tool. Pay per token, no subscription.
Moonshot's launch post says it is available in chat and agent mode at kimi.com.
Download the open weights from Hugging Face. Needs a multi-GPU server and the skills to run it.
On the API you pay for the text going in and the text coming out. Enter a month's usage to compare the two versions.
List prices before tax. "Repeated input" is text the model has already seen, which costs much less. Coding tools resend a lot of the same context, so the share is often high, but yours will differ. Thinking counts as output.
Coding every day? A Kimi membership with Kimi Code may cost less than paying per token. See API vs subscription, API pricing and Kimi AI pricing.
A coding model works best with a clear goal and a way to check its work. Replace the parts in [brackets].
The test [test name] is failing. Find the cause, fix it with the smallest change you can, and run the tests again to confirm.
Read this project and explain how it is organised. List the main folders, what each does, and where I should look to change [feature].
Add [feature] to this project. First show me your plan and wait for my OK. Then make the changes and add tests for them.
Full walkthrough: Kimi Code setup and first coding task.
| Kimi K2.7 Code | Kimi K2.6 | Kimi K3 | |
|---|---|---|---|
| Made for | Programming | Everyday general work | The hardest tasks of any kind |
| Reads at once | About 262K tokens | About 262K tokens | About 1 million tokens |
| Thinking | Always on | On or off | Always on |
| API price, in / out | $0.95 / $4.00 | $0.95 / $4.00 | $3.00 / $15.00 |
| Faster version | Yes, HighSpeed | No | No |
| Pick it for | Daily coding at a low price | Chat, writing and mixed tasks | Huge codebases and deep reasoning |
On a small screen, swipe the table sideways to see every column.
More: Kimi K2.6 · Kimi K3 · choose a Kimi model