Qwen3.8 27B GGUFs exhibit statistically lossless performance compared to FP8, according to user report

IDK if it's related to https://arxiv.org/abs/2605.02404 https://github.com/IST-DASLab/SLQ but I dl'd the EAR99 on a whim & like a very unscientific 30k tokens later I can confirm it feels same or better as FP8 :) ...