Kimi K2.7 Code

Moonshot's 1T-parameter coding agent model: 32B active, 256K context, native INT4, image and video input, modified MIT license.

In short. Kimi K2.7 Code is a very large AI model from Moonshot AI built for software engineering. It can work through long coding tasks with many tool calls, and it also reads images and video. The license is the MIT license with one extra sentence that only affects very large products.

Summary

Moonshot AI released Kimi K2.7 Code on 12 June 2026. It is a coding-focused agentic model built on Kimi K2.6: a mixture-of-experts model with about 1 trillion parameters, of which 32 billion are active per token. It has 61 layers (one of them dense), 384 experts with 8 chosen per token plus one shared, multi-head latent attention and a 256K-token context window. A 400-million-parameter MoonViT encoder adds image and video input. The weights are shipped natively in INT4 and take about 595 GB.

Moonshot says the release has higher success rates on long-horizon coding, on sequences of 4,000 or more tool calls and 12 or more hours of continuous execution, and cuts thinking-token use by roughly 30 percent compared with K2.6. Thinking mode is always on, and the recommended sampling is temperature 1.0 and top-p 0.95. Its own table, which is self-reported, shows Kimi Code Bench v2 at 62.0 (K2.6: 50.9), Program Bench 53.6, MCP Atlas 76.0 and MCP Mark Verified 81.1. K2.7 Code beats K2.6 on every row but trails GPT-5.5 and Claude Opus 4.8 on most of them, and Kimi Code Bench is Moonshot's in-house benchmark.

The model card recommends vLLM, SGLang or KTransformers, and the same deployment recipe as K2.5 and K2.6, with transformers 4.57.1 or later but below 5.0. Video input is described as experimental and works only through Moonshot's own API for now. The license is a Modified MIT License: use, modification and sale are allowed, and only a product with over 100 million monthly active users or over US$20 million of monthly revenue must show "Kimi K2.7 Code" in its interface.

Good for

  • Self-hosted coding agents that run long tool-call sequences
  • Software work that mixes screenshots or video with code, such as UI bugs
  • Companies that want an almost-MIT license for a trillion-parameter model
  • Teams already using Kimi Code CLI, which Moonshot recommends as the agent framework

Watch out for

  • It needs a server-class GPU node, since even the INT4 weights are about 595 GB
  • Thinking cannot be turned off, and the official API fixes the sampling settings
  • Video input works only through Moonshot's own API for now
  • The benchmark numbers come from Moonshot, including its in-house Kimi Code Bench

Performance claims come from the developer and have not been independently verified.

How much memory does it need?

Estimates for the K2.7 Code (INT4) variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.

Variants Parameters BF16 FP8 / INT8 4-bit (Q4) Fits on
K2.7 Code (INT4) INT4 1.03T 32B active 2.36 TB 1.18 TB 655 GB GPU server

Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.

VRAM calculator

The license in plain words

The standard MIT license with one added sentence. Anyone can use, modify and sell the model, and only very large products (over 100 million monthly active users or over US$20 million of monthly revenue) must show the model's name in their interface.

  • Attribution or model name must be shown

More about Modified MIT

See also

Frequently asked questions

What is Kimi K2.7 Code?
Kimi K2.7 Code is a very large AI model from Moonshot AI built for software engineering. It can work through long coding tasks with many tool calls, and it also reads images and video. The license is the MIT license with one extra sentence that only affects very large products.
Can I use Kimi K2.7 Code commercially?
Yes, with conditions. Kimi K2.7 Code is released under the Modified MIT, which allows commercial use but adds extra rules: Attribution or model name must be shown. Read the license text before you ship a product.
Is Kimi K2.7 Code open source?
Kimi K2.7 Code is open-weight, not open source in the strict sense. You can download the model files, but the Modified MIT adds conditions that the Open Source Initiative definition does not allow.
Can I run Kimi K2.7 Code on my own computer?
Not on a desktop. The smallest variant, K2.7 Code (INT4), needs about 655 GB at 4-bit precision, which means a GPU server with several data-center cards.
When was Kimi K2.7 Code released?
Kimi K2.7 Code was released by Moonshot AI on June 12, 2026.
How long a context window does Kimi K2.7 Code have?
Kimi K2.7 Code supports a context window of up to 256K tokens.

Official links

Sources

Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.

Release date and facts last verified November 11, 2026.

Related models

Model Released Parameters Context License Smallest variant needs
Kimi K3 Moonshot AI 2.8T 104B active 1M Kimi K3 License (Conditional) 1.72 TB Multi-node cluster
Kimi K3.1 Pro Moonshot AI 2.8T 104B active 1M Kimi K3 License (Conditional) 1.72 TB Multi-node cluster
Kimi K2.6 Moonshot AI 1.03T 32B active 256K Modified MIT (Conditional) 655 GB GPU server
GLM-5.2 Z.ai (Zhipu AI) 753B 40B active 1M MIT (Permissive) 520 GB GPU server