Reddit user compares Qwen 3.6 Q4 vs Gemma 4 Q8 for local code tasks

User with 32GB unified memory gets ~15 tok/s on Qwen 3.6 35B-A3B @ Q4 and asks if Gemma 4 12B @ Q8 is better for coding. Discussion covers quantization trade-offs and practical performance.