Gemma 4 2B runs well on old hardware, user reports

A Reddit user reports running Gemma 4 2B at 9 t/s on an i5 6500, finding output better than ChatGPT 3.5 and comparable to 4. They also praise Qwen 3.5 4B as another good small model.