TurboFieldfare - Run Gemma 4 26B on M-series Macs with 2 GB RAM
Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac

TurboFieldfare is a custom Swift and Metal runtime that enables the Gemma 4 26B-A4B model to run on any Apple Silicon Mac, even those with only 8 GB of RAM. By keeping the shared core in memory and streaming expert weights directly from the SSD, it achieves inference with a mere 2 GB memory footprint. This open-source project offers a native Mac app, a CLI, and an OpenAI-compatible server, allowing developers to leverage powerful large language models without expensive hardware upgrades. Built specifically for macOS 26 and Metal 4, it bypasses generic wrappers to deliver optimized performance for local AI tasks.
Memory got expensive. So I gave a 26-billion-parameter model a ~2 GB budget.