Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac
8.1 relevance
Score Breakdown
technical depth 9
novelty 8
actionability 8
community 8
strategic 5
personal 9
Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.
Open-source inference engine for running Gemma 4 on Mac with minimal RAM is highly relevant to AI/ML deployment and developer experience.
Summary
TurboFieldfare is an open-source Swift + Metal runtime that runs Google's Gemma 4 26B-A4B model on any Apple Silicon Mac using only ~2 GB of RAM by streaming expert weights from SSD instead of loading the full 14.3 GB model. It achieves 5.1-6.3 tok/s on an 8 GB M2 MacBook Air and 31-35 tok/s on a 24 GB M5 Pro, with a native Mac app, CLI, and OpenAI-compatible server. The engine is model-specific, not a wrapper around MLX or llama.cpp, and requires macOS 26 with Metal 4 and Swift 6.2.
Author
drumih