# google/gemma.cpp

lightweight, standalone C++ inference engine for Google's Gemma models.

Repository: https://github.com/google/gemma.cpp
Canonical: https://ross.abutalabs.com/products/gemmacpp
Language: C++
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T18:54:38+00:00

## Health v2 (maintenance only)
Score: 72/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 40, longevity 66
- inputs: {"age_days": 932, "days_push": 7, "days_rel": 526, "gap_med": 11, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7031, forks 659 (observed 2026-08-28T04:09:55.062486+00:00)

## What it is
A lightweight, standalone C++ inference engine for Google's Gemma foundation models (Gemma 2/3, PaliGemma 2), with a small ~2K LoC core and portable SIMD via Google Highway. It is designed for experimentation and research rather than production deployment.

## Use cases
- run gemma models locally on cpu
- embed an llm inference engine in a c++ project
- experiment with llm model internals and low-level computation
- research on quantized weight compression and mixed-precision gemm
- run a small language model without python dependencies
- modify an llm runtime for algorithm research

## When to choose
- you want a minimal, hackable C++ codebase for LLM research
- you need CPU-only inference of Gemma models with minimal dependencies
- you want to embed Gemma inference directly into another C++ application
- you want to experiment with SIMD, quantization, or backward passes on transformer models

## When to avoid
- you need production-grade edge or server LLM deployment - use JAX, PyTorch, or Transformers instead
- you need GPU acceleration or broad multi-model support - use llama.cpp or vLLM
- you want a turnkey chat application rather than an inference library

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, deep-learning
- domain: large-language-models, artificial-intelligence, machine-learning
- platform: cpp, windows
- tags: gemma, cpu-inference, simd, on-device-llm, minimalist, research, linux, macos, cpu

## Member repositories
- google/gemma.cpp (main) score 72

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:55.062486+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 2026-08-29T17:40:35.160085+00:00, confidence not recorded.
  - readme: https://github.com/google/gemma.cpp (fetched 2026-08-28T04:09:55.062486+00:00, sha 66ffd8778b1d)
- Data as of 2026-08-30T08:39:29.467469+00:00.