#53 · Primary category： Inference & Local Deploy
turbo-fieldfare
Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook
Project last updated：08/29/26
GitHub Stars
6.5K
Forks
408
Contributors
1
License
Apache-2.0
Why we included this project
Anyone with an 8 GB Mac and a long-standing wish to run a real 26-billion-parameter model locally will find a lot to like here. TurboFieldfare is a Swift and Metal runtime written specifically for Gemma 4 26B-A4B, not a thin wrapper around MLX or llama.cpp. It keeps the model's shared core and KV cache resident and pulls the routed experts in from SSD on demand, which is how it fits roughly 2 GB of memory and still works on base Apple Silicon laptops. You get a native Mac app, a CLI, and an OpenAI-compatible local server, so existing chat-completions tooling can talk to it directly. The project also documents its design and more than a hundred measured experiments, a level of transparency that makes it easier to trust the performance claims before you commit.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
ollama
Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
llama.cpp
LLM inference in C/C++
vllm
A high-throughput and memory-efficient inference and serving engine for LLMs
gpt4all
GPT4All: Run Local LLMs on Any Device. Open-source and available for commercial use.
LocalAI
LocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.