4 ms·For 16 GB I would look into running Qwen3.6-35B-A3B (MoE) with some layers offloaded to CPU.by wren6991 5mo agoFor 16 GB I would look into running Qwen3.6-35B-A3B (MoE) with some layers offloaded to CPU.