Qwen3.8 125B (6B active) Flash-Next

User reports Qwen3.8-Flash-Next at 40 t/s on an M5 Mac with 128GB unified memory. Setup is oMLX runtime, model fits in 90-95GB of RAM. User notes MTPLX reaches 60 t/s with more tuning, and Qwen3.6 MoE on Splash runtime hits 120 t/s for most tasks except coding.