I tested Qwen3.8 27B KLD for 21 Models at 9 different KV quants that fit on 16GB VRAM (189 benchmarks).

🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1x6fwcd/i_tested_qwen38_27b_kld_for_21_models_at_9/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-15 16:11:09
Reddit
From the LocalLLaMA community on Reddit
From the LocalLLaMA community on Reddit
Explore this post and more from the LocalLLaMA community