TurboFieldfare
A custom Swift and Metal runtime that runs the 26-billion-parameter Gemma 4 26B-A4B model in about 2 GB of RAM on any Apple Silicon Mac, including 8 GB models.
TurboFieldfare avoids loading the full 14.3 GB Gemma 4 26B-A4B model into memory by keeping a shared 1.35 GB core and an FP16 KV cache resident, then streaming only the experts a given token needs from disk. This is what lets the model run on Macs with as little as 8 GB of RAM, at a measured 5 to 6 tokens per second on an 8 GB M2 MacBook Air and 31 to 35 tokens per second on a 24 GB M5 Pro.
It is model-specific rather than a general wrapper around MLX or llama.cpp, and its curated experiment record summarizes 103 measured results across kernel, caching, I/O, prefill, and decode changes. The release build includes a foreground Mac app and a companion decode-service executable, built with Swift Package Manager.
It requires macOS 26, Metal 4, and Swift 6.2, and is distributed under the Apache-2.0 license.