AnalysisAI ModelsAugust 28, 2026

Reddit user touts Qwen 3.8 27B QAT Q2 GGUF quant

A LocalLLaMA post recommends running Qwen 3.8 27B at Q2 quantization with QAT weights, paired with a Q2 DFlash model and Q5 KV cache. Two GGUF repos are linked: sdkyuan/qwen3.8-27B-qat-q2_0-gguf and HermiHg/Qwen3.8-27B-DFlash2-Q2_K_S-MIX-GGUF.

1 source

More stories today

Open the live feed