Qwen3.8-27B FP8 quantization quality questioned by users

Users report that Qwen3.8-27B's official FP8 quantization shows worse KL divergence than Q80 despite similar size, sparking debate over quantization choices including unsloth's UD-Q8-K-XL GGUF for long-horizon agent tasks.

2026-08-18 ~ 2026-08-18 · 2 related posts