# google-deepmind/gemma

Gemma open-weight LLM library, from Google DeepMind

Repository: https://github.com/google-deepmind/gemma
Canonical: https://ross.abutalabs.com/products/gemma
Homepage: https://gemma-llm.readthedocs.io
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T04:08:08+00:00

## Health v2 (maintenance only)
Score: 87/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 84, longevity 66
- inputs: {"age_days": 925, "days_push": 7, "days_rel": 105, "gap_med": 1, "n_releases_24m": 18}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5695, forks 1014 (observed 2026-08-28T04:09:28.015850+00:00)

## What it is
The official JAX-based Python library from Google DeepMind for running, sampling from, and fine-tuning the Gemma family of open-weight large language models. It supports multi-turn and multi-modal chat, LoRA fine-tuning, and checkpoint loading across Gemma versions 2 through 4.

## Use cases
- run gemma llm locally with jax
- fine-tune gemma model
- lora fine-tuning for gemma
- multi-turn chat with gemma
- multimodal image chat with gemma
- load gemma checkpoints in python

## When to choose
- you want to use or fine-tune Gemma models in JAX on CPU, GPU, or TPU
- you need the official DeepMind implementation with multi-modal and multi-turn chat support
- you want LoRA-based parameter-efficient fine-tuning of Gemma

## When to avoid
- you need PyTorch or transformers-based tooling instead of JAX
- you want to run other model families beyond Gemma
- you need a production serving stack rather than a research library

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, machine-learning, chatbot
- domain: large-language-models, deep-learning, artificial-intelligence
- platform: python, cloud
- tags: jax, open-weights, fine-tuning, lora, multimodal, google-deepmind, gpu

## Member repositories
- google-deepmind/gemma (main) score 87

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:28.015850+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 2026-08-29T17:53:38.235017+00:00, confidence not recorded.
  - readme: https://github.com/google-deepmind/gemma (fetched 2026-08-28T04:09:28.015850+00:00, sha 7f774bd3b896)
  - registry_pypi: https://pypi.org/pypi/gemma/json (fetched 2026-08-29T08:49:36.744286+00:00, sha 15804cf1cd45)
- Data as of 2026-08-30T08:39:29.467469+00:00.