Qwen3.8 125B (6B active) Flash-Next
on Unknown GPU
Summary
User reports running a model on a 12 GB mid-range Android phone with optimizations and low quantization on the dense part.
The model size is given as 80 GB, likely the total including quantized weights.