# MoonshotAI/Kimi-Dev

open-source coding LLM for software engineering tasks

Repository: https://github.com/MoonshotAI/Kimi-Dev
Canonical: https://ross.abutalabs.com/products/kimi-dev
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-09-30T02:15:54+00:00

## Health v2 (maintenance only)
Score: 38/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 44, release rhythm 35, longevity 31
- inputs: {"age_days": 443, "days_push": 338, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1335, forks 186 (observed 2026-08-28T04:04:25.194964+00:00)

## What it is
Kimi-Dev-72B is an open-source coding LLM from Moonshot AI specialized in software engineering tasks like issue resolution and bug fixing, achieving state-of-the-art results on SWE-bench Verified among open models. The repository provides tooling to deploy the model with vLLM and run its two-stage file-localization and patch-generation pipeline on real repositories.

## Use cases
- resolve github issues automatically with an llm
- patch real repositories to fix bugs
- benchmark coding llms on swe-bench verified
- self-host a 72b coding model with vllm
- generate unit tests for code repair tasks
- research rl training for software engineering agents

## When to choose
- you need an open-weights model tuned for issue resolution and repo-level bug fixing
- you want to reproduce or compare SWE-bench Verified results
- you have multi-GPU infrastructure to self-host a 72B model

## When to avoid
- you need a lightweight coding assistant without large GPU resources
- you want a general-purpose chat model rather than a specialized repair pipeline
- you need a hosted API service rather than self-deployment

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, agent-framework
- domain: large-language-models, developer-tools, artificial-intelligence
- platform: python
- tags: coding-llm, swe-bench, issue-resolution, bug-fixing, reinforcement-learning, vllm, 72b-model, open-weights, docker, gpu, linux

## Member repositories
- MoonshotAI/Kimi-Dev (main) score 38

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:25.194964+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:44:21.270578+00:00, confidence not recorded.
  - readme: https://github.com/MoonshotAI/Kimi-Dev (fetched 2026-08-28T04:04:25.194964+00:00, sha 235102fd7316)
- Data as of 2026-08-30T08:39:29.467469+00:00.