# ise-uiuc/magicoder

[ICML'24] Magicoder: Empowering Code Generation with OSS-Instruct

Repository: https://github.com/ise-uiuc/magicoder
Canonical: https://ross.abutalabs.com/products/magicoder
Homepage: https://proceedings.mlr.press/v235/wei24h.html
Language: Python
License: MIT
License Family: permissive
Topics: ai4code, large-language-models, llm, llm4code
Last push: 2024-11-01T17:28:52+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 73
- inputs: {"age_days": 1027, "days_push": 670, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2094, forks 170 (observed 2026-08-28T04:06:13.150348+00:00)

## What it is
Magicoder is a family of fully open-source code LLMs (under 7B parameters) trained with OSS-Instruct, a method that seeds LLMs with open-source code snippets to generate diverse, low-bias instruction data. The repository provides training code, datasets, and model checkpoints released under MIT license.

## Use cases
- generate synthetic instruction data for code LLMs
- fine-tune a small code generation model
- reproduce OSS-Instruct training pipeline
- find open weights for a code assistant model
- compare small code models on HumanEval
- build a self-hosted code completion model

## When to choose
- you need fully open code, weights, and data for a code LLM
- you want to mitigate bias in LLM-synthesized instruction data
- you need a sub-7B code model competitive with larger ones
- you want a proven recipe cited by CodeGemma and OpenCodeInterpreter

## When to avoid
- you need a production coding assistant with product support
- you require models larger than 7B for frontier performance
- you lack GPU resources for training or inference
- you need a general-purpose (non-code) LLM

## Facets
- artifact type: learning-resource
- maturity: stable
- function: llm-training, machine-learning, data-generation
- domain: large-language-models, machine-learning, developer-tools
- platform: python
- tags: code-generation, oss-instruct, instruction-tuning, synthetic-data, icml24, open-source-models, gpu

## Member repositories
- ise-uiuc/magicoder (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:13.150348+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 2026-08-30T02:55:25.510929+00:00, confidence not recorded.
  - readme: https://github.com/ise-uiuc/magicoder (fetched 2026-08-28T04:06:13.150348+00:00, sha 88ac2c63d088)
  - homepage: https://proceedings.mlr.press/v235/wei24h.html (fetched 2026-08-29T10:35:06.098160+00:00, sha c9d2deb83979)
  - site_page: https://proceedings.mlr.press/faq.html (fetched 2026-08-29T10:35:06.100842+00:00, sha 8708020cd335)
- Data as of 2026-08-30T08:39:29.467469+00:00.