# ibm-granite/granite-code-models

Granite Code Models: A Family of Open Foundation Models for Code Intelligence

Repository: https://github.com/ibm-granite/granite-code-models
Canonical: https://ross.abutalabs.com/products/granite-code-models
Homepage: https://huggingface.co/collections/ibm-granite/granite-code-models-6624c5cec322e4c148c8b330
License: Apache-2.0
License Family: permissive
Archived: true
Last push: 2025-06-25T20:34:38+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 28, release rhythm 35, longevity 61
- inputs: {"age_days": 862, "days_push": 434, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1251, forks 85 (observed 2026-08-28T04:04:08.141902+00:00)

## What it is
A family of open decoder-only code foundation models from IBM, released under Apache 2.0 in base and instruct variants at 3B, 8B, 20B, and 34B parameter sizes. The repository serves as the official hub with links to the paper and Hugging Face model collection, describing training data and capabilities for code generation, explanation, fixing, and translation.

## Use cases
- generate code from natural language instructions
- explain or document existing code
- fix bugs automatically with an LLM
- translate code between programming languages
- self-host an open code model for enterprise use
- fine-tune a code LLM on private repositories

## When to choose
- you need a permissively licensed (Apache 2.0) code LLM for commercial use
- you want enterprise-grade models trained on license-clean data with PII redaction
- you need a range of model sizes to trade off compute vs capability
- you want one model covering generation, explanation, fixing, and translation

## When to avoid
- you need the absolute highest coding benchmark scores regardless of license
- you cannot host multi-billion-parameter models locally
- you want a hosted API rather than open weights
- you need multimodal or non-code general-purpose capabilities

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-inference, machine-learning, nlp
- domain: large-language-models, developer-tools, artificial-intelligence
- platform: python, cross-platform
- tags: code-llm, foundation-models, ibm-granite, apache-2.0, huggingface, code-generation, open-weights, gpu

## Member repositories
- ibm-granite/granite-code-models (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:08.141902+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 2026-08-30T05:07:39.973682+00:00, confidence not recorded.
  - readme: https://github.com/ibm-granite/granite-code-models (fetched 2026-08-28T04:04:08.141902+00:00, sha 2273f9807e84)
- Data as of 2026-08-30T08:39:29.467469+00:00.