Tested in Coding: Q8_K_XL Qwen3.8 27B vs BF16 Qwen3.6 27B
英文标题：Tested in Coding: Q8_K_XL Qwen3.8 27B vs BF16 Qwen3.6 27B
Further to my last post, https://www.reddit.com/r/LocalLLaMA/comments/1vldngi/tested_in_coding_bf16_muse_glimmer_vs_bf16_qwen36/, I bring this requested comparison of Qwen3.8 27B. Context: Both models are running at full FP16 KV-cache. Due to RAM limitations, BF16 Qwen3.6 was running at 150,000, whe
打开原文相关 4 · 机会 0
中文
Further to my last post, https://www.reddit.com/r/LocalLLaMA/comments/1vldngi/tested_in_coding_bf16_muse_glimmer_vs_bf16_qwen36/, I bring this requested comparison of Qwen3.8 27B. Context: Both models are running at full FP16 KV-cache. Due to RAM limitations, BF16 Qwen3.6 was running at 150,000, whe
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
Further to my last post, https://www.reddit.com/r/LocalLLaMA/comments/1vldngi/tested_in_coding_bf16_muse_glimmer_vs_bf16_qwen36/, I bring this requested comparison of Qwen3.8 27B. Context: Both models are running at full FP16 KV-cache. Due to RAM limitations, BF16 Qwen3.6 was running at 150,000, whe