Statistically Lossless Qwen3.8 27B GGUF Quants Hit ~99% EAR at 23GB, Feels FP8-like
Dany0 · reddit · 2026-10-28
A Reddit user tested new "statistically lossless" GGUF quantizations of Qwen3.8 27B, possibly from the IST-DASLab SLQ paper (arXiv 2605.02404). After 30k tokens of informal testing, they report quality on par with or better than FP8.
A fidelity-vs-size chart shows SLQ EAR99 (23.3GB) reaching 99% EAR, clearly beating the similarly-sized UD-Q6KXL (24.9GB), with EAR98.5 and EAR98 checkpoints also outperforming donor GGUFs at matching sizes. If it holds up, this means near-full-precision quality at lower VRAM for local deployments.
Related event: Statistically Lossless Quantized GGUF of Qwen3.8 27B Released(3 posts)→
More from Infra
- EmbeddingGemma 2 runs real-time embeddings on a MacBook CPU alone: 20ms text, 100ms images — generativist · 2026-10-28
- Statistically lossless Qwen3.8-27B GGUF quantizations land on Hugging Face, analysis paywalled — Dany0 · 2026-10-28
- Nebius benchmark: 8 small GPU replicas yield 42.5% more Cosmos video per hour — shashib · 2026-10-28
- Data centers become the new front line as AI, cyber, physical and OT threats converge — ChuckDBrooks · 2026-10-28
- preowned turns macOS built-in Apple models into an OpenAI-compatible local API — pcuenq · 2026-10-28
- 2,100+ tok/s/user for DeepSeek V4.1 Flash on 8× AMD MI355X, 2.8x faster than open-source — JiaZhihao · 2026-10-28