Skip to content

Kimi-Dev

An open-source coding LLM for software engineering tasks

Visit official site

What Kimi-Dev is

The model is described as an open-source coding LLM aimed at software engineering problems, optimised with large-scale reinforcement learning for correctness and robustness in real development settings.

What you can do with it

  • Locate and fix code errors from a problem description
  • Generate unit tests for existing code
  • Tune repair results with reinforcement learning
  • Deploy the model locally or in the cloud

Model notes

  • Kimi-Dev-72B is reported at 60.4% on SWE-bench
  • The model is optimised with large-scale reinforcement learning
  • The material names Hugging Face and GitHub for downloads and deployment

Getting started

  • Clone the Kimi-Dev repository with git clone
  • Create and activate a virtual environment with conda
  • Install required dependencies with pip install
  • Download and extract the processed data for the project structure

Pros & cons

✓ What we like

  • Open source with community contributions encouraged
  • Deployment is described for both local and cloud settings

! What to watch out for

  • Deployment steps start with cloning the repository and setting up an environment
  • The reported 60.4% figure is tied to the Kimi-Dev-72B variant

FAQ

What number is reported for Kimi-Dev-72B?

The material reports 60.4% on SWE-bench.

Where can the model be downloaded?

The material lists Hugging Face and GitHub for downloads and deployment.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review