Samosa Chat runs Qwen3 30B-class model locally on a 16 GB Mac

Samosa Chat lets you run Qwen3.6-35B-A3B — a mixture-of-experts model that activates roughly 3B parameters per token — directly on a 16 GB MacBook, no cloud required. For indie makers who want capable local AI without paying API bills or sending data offsite, squeezing a model this size into consumer RAM is a meaningful practical step.