Qwen3.8-Flash-Next on 2x RTX 5090 with 504 tok/s decode: QEngine 1.3.1 decodes ~1.5-1.9x faster than the next-best engine I've tried (numbers, caveats, and what still bites)
https://github.com/bpx03/QEngine (not my repo, credit goes to bpx03) QEngine was built for 3x RTX 3090s. I was able to compile for RTX 5090s. I've now run Qwen3.8-Flash-Next on the same box through four different engines.