TurboFieldfare
TLDR Dev
TurboFieldfare is a Swift and Metal runtime that runs Google's Gemma 4 26B language model in approximately 2 GB of RAM on Apple Silicon Macs by streaming expert weights from SSD rather than loading the full 14.3 GB model into memory. The system achieves 5.1–6.3 tokens per second on an 8 GB M2 MacBook Air and 31–35 tokens per second on a 24 GB M5 Pro by keeping only a 1.35 GB core and KV cache resident while fetching needed experts on-demand. Users can now run a 26-billion-parameter instruction-tuned model natively on entry-level Macs through a native app, CLI, or OpenAI-compatible server.
Why it matters
TurboFieldfare is a custom runtime designed for inference of the Gemma 4 26B-A4B model on Apple Silicon Macs with limited RAM, allowing the model to operate efficiently without loading the entire 14.3 GB into memory. The system uses streaming and caching techniques to maintain performance while only using approximately 2 GB of RAM.
