Qwen-Next seems worse to me then 3.8 27b for coding, but I feel like I must be missing something?
一位用户在MTPLX上运行Qwen-Next和3.8 27b模型进行编码任务，特别是在nodejs的pi agent中。尽管拥有128GB内存并运行q8 27b模型（可能由于“optimized for speed”设置而以动态q4和8位注意力运行），但用户发现3.8 27b在处理更复杂的任务时表现更强。尽管许多人认为Qwen-Next更好，但用户的实际体验却相反，这让他们怀疑自己是否理解有误。
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年9月12日 00:05 UTC
收录当时偏移：UTC+02026年9月12日 15:01 UTC
- 发布
- 2026年9月12日 00:05
- 收录
- 2026年9月12日 15:01
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
Hi! I run both models on MTPLX on my m5 max, and since I have 128GB of ram I run the q8 27b. I think MTPLX only lets me run "optimized for speed" which it says is a dynamic q4 with 8 bit attention.
Both of them honestly are very speedy! For coding (in pi agent in nodejs) I've just noticed that 27B feels stronger with harder tasks. But I've read so many people on here say qwen-next is better so I was wondering if maybe I'm just doing or thinking about it wrong?
(and p.s. its sooo amazing that alibaba just made and released this amazing models for free! ❤️)