# OpenCoder-llm/OpenCoder-llm

The Open Cookbook for Top-Tier Code Large Language Model

Repository: https://github.com/OpenCoder-llm/OpenCoder-llm
Canonical: https://ross.abutalabs.com/products/opencoder-llm
Homepage: https://opencoder-llm.github.io/
Language: Python
License: MIT
License Family: permissive
Last push: 2024-12-08T16:46:00+00:00

## Health v2 (maintenance only)
Score: 22/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 48
- inputs: {"age_days": 676, "days_push": 633, "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 2111, forks 126 (observed 2026-08-28T04:06:14.303240+00:00)

## What it is
OpenCoder is a fully open and reproducible family of code large language models (1.5B and 8B base and chat variants) trained on 2.5 trillion tokens, released with model weights, training data, data processing pipelines, and evaluation frameworks. The repository serves as the central hub ('open cookbook') for the models, the RefineCode pretraining corpus, SFT datasets, intermediate checkpoints, and the OpenCodeEval evaluation framework.

## Use cases
- train a code large language model from scratch
- download open code llm weights for code generation
- build a reproducible code pretraining corpus
- evaluate code llms on benchmarks like humaneval
- study ablation results for code llm training strategies
- fine-tune a code model with open sft datasets
- clean and filter code data for pretraining

## When to choose
- you need a fully open, reproducible code LLM with transparent data pipeline
- you want to reproduce or extend top-tier code model training
- you need an efficient evaluation framework for code generation models
- you want high-quality open code pretraining and SFT datasets

## When to avoid
- you just need a hosted code assistant without training infrastructure
- you need the largest closed models like GPT-4 or Claude for coding
- you lack GPU resources for running or training multi-billion parameter models
- you need a production inference server rather than research artifacts

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-training, llm-inference, machine-learning, etl, benchmarking, data-generation
- domain: large-language-models, machine-learning, deep-learning, developer-tools
- platform: python, cross-platform
- tags: code-llm, open-weights, pretraining, refinecode, opencodereval, huggingface, reproducible-research, sft-dataset, data-engineering, gpu, linux

## Member repositories
- OpenCoder-llm/OpenCoder-llm (main) score 22

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:14.303240+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 2026-08-30T02:53:42.577109+00:00, confidence not recorded.
  - readme: https://github.com/OpenCoder-llm/OpenCoder-llm (fetched 2026-08-28T04:06:14.303240+00:00, sha ad4cd5c3696a)
  - homepage: https://opencoder-llm.github.io/ (fetched 2026-08-29T10:33:44.132596+00:00, sha c8d149227395)
- Data as of 2026-08-30T08:39:29.467469+00:00.