Community quants for Qwen3.8 Flash save 20-30GB at same quality as unsloth
Dutchnamn · reddit · 2026-08-29
A Reddit user released a set of Qwen3.8 Flash (Next) GGUF quants after days of benchmarking. They require 20-30GB less disk and RAM than comparable unsloth or AesSedai quants at equal quality. PPL is published in the readme and is competitive; both Q4 quants are strong, with Q3/Q5 to follow. Built with imatrix and a tailored quantization recipe. A ROCmFP4 variant for AMD users is slightly better and faster than Q4XS.
More from Infra
- Meta open-sources Muse Spark Open: 816B-parameter model that runs locally on two Mac Studios — gabriberton · 2026-10-16
- Jensen Huang joins BG2 podcast: does AI revenue justify capex, '27 bottlenecks — elonmusk · 2026-10-16
- Mistral's on-device infra rack revealed: DGX Spark, Mac Studio and more — gregmushen · 2026-10-16
- NVIDIA and Sweden's Wallenberg launch Sferical One, a shared AI supercomputer — nvidia · 2026-10-16
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16
- Xiaomi's MiMo RL dashboard reveals $3.47M spent on just 30 training steps — bycloud · 2026-10-16