Qwen 3.6 27B + Opencode: what am i doing wrong?
英文标题：Qwen 3.6 27B + Opencode: what am i doing wrong?
Ok i'm sort of getting desperate here. So i am currently running Qwen 3.6 27B locally at Q8_K_XL and with only ~105k CTX at q8_0. Before i tried it at Q8_0 but with F16 KV cache. Of course i also tried lower quants down to Q6 and bigger context sizes up to 256k at F16 For a harness i use Opencode wi
打开原文相关 2 · 机会 0
中文
Ok i'm sort of getting desperate here. So i am currently running Qwen 3.6 27B locally at Q8_K_XL and with only ~105k CTX at q8_0. Before i tried it at Q8_0 but with F16 KV cache. Of course i also tried lower quants down to Q6 and bigger context sizes up to 256k at F16 For a harness i use Opencode wi
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
Ok i'm sort of getting desperate here. So i am currently running Qwen 3.6 27B locally at Q8_K_XL and with only ~105k CTX at q8_0. Before i tried it at Q8_0 but with F16 KV cache. Of course i also tried lower quants down to Q6 and bigger context sizes up to 256k at F16 For a harness i use Opencode wi