On: Qwen3 30B A3B Hits 13 token/s on 4xRaspberry Pi 5

[fastball]

Capability of the model itself is presumably the more important question than those other two, no?