Qwen 3 on a Raspberry Pi 5: Small Models, Big Agent Energy

TL;DR – We ran the brand-new Qwen 3 (0.6 B & 1.7 B) on our Distiller CM5 dev-kit.≈9 tokens/s on the 1.7 B Q8 build and 21 tokens/s on the 0.6 B Q8 build—both under 1.3 GB RAM, and—more interesting—noticeably “agentic” behaviour that auto-chains Wi-Fi...