I ported TurboFieldfare to Qwen 3.6 35B and it runs in 1.4 GB of RAM

Was playing around with TurboFieldfare, a Mac engine that runs Gemma 4 26B in ~2 GB by streaming MoE experts off SSD instead of loading them. It only supported that one model, so I added support for Qwen 3.6 35B-A3B. Comparatively, Qwen needs lesser memory. ~1.4 GB vs ~2.1 GB for Gemma. Qwen's

中文全文翻译正在补齐，可先查看下方英文原文。

Was playing around with TurboFieldfare, a Mac engine that runs Gemma 4 26B in ~2 GB by streaming MoE experts off SSD instead of loading them. It only supported that one model, so I added support for Qwen 3.6 35B-A3B. Comparatively, Qwen needs lesser memory. ~1.4 GB vs ~2.1 GB for Gemma. Qwen's