drumih/turbo-fieldfare
Turbo Fieldfare enables low-memory Gemma 4 26B inference on Apple Silicon, broadening local LLM deployment on consumer Macs.
Excerpt
Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook — ★ 522 · Swift · topics: apple-silicon, gemma, gemma4, gemma4-26b-a4b, gpgpu, llm
Read at source: https://github.com/drumih/turbo-fieldfare