For developers who live in coding agents, it promises fewer thinking tokens than K2.6 at the same price, with open weights too.

Kenning's take

What it does

Kimi K2.7 Code is a coding-focused version of Kimi K2.6 for long software engineering tasks. Moonshot says it uses about 30% fewer thinking tokens than K2.6 while scoring higher on its own coding benchmarks. It runs best with the Kimi Code command-line tool, is sold on the Kimi API in standard and high-speed versions, and has open weights on Hugging Face.

• Built on Kimi K2.6 for coding agents
• About 30% fewer thinking tokens than K2.6
• 256K context with image and video input
• High-speed API variant at about 180 tokens per second
• Modified MIT licence weights

Who it suits. Developers who use coding agents and want an open model they can call by API or host.

Key facts

Category

AI models

Part of

Kimi

Made by

Moonshot AI

Based in

China

Released

12 June 2026

How to get it

on the Kimi API (kimi-k2.7-code and kimi-k2.7-code-highspeed), in the Kimi Code CLI, and as open weights on Hugging Face

Works on

API, CLI, self-hosted

Free plan

Weights free to download

API price

$0.95 per million input tokens ($0.19 cached), $4 per million output; highspeed $1.90 input, $8 output

Reviews and write-ups

• Handy AI (Substack) 2026-06-12

FAQ

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.