# Observations from  seeing Qwen 4 max’s outputs — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/961ae9a398e28a1b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-11-11 20:43 ingested 2026-11-12 02:42

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xtt3rb/observations_from_seeing_qwen_4_maxs_outputs/)

## retained text



```
I thought about not writing this but i feel like i should.. from the  info I gathered,  Qwen 4 max is slightly worse than opus 5.5 medium… Meaning it is definitely  above opus low which is  hy4 level and opus 5.5 low is  above ds v4.1 flash max.  wow, it means it is somewhere between  hy4/ds v4.1 flash 1017 max   and kimi 3.1 pro max . I thought qwen 4 max would do a little better. i guess another 4-6  months until we get an open or Chinese opus 5.5  max model..

This should mean  qwen 4 27b will be worse than ds v4.1 flash 1017  max but better than mistral 4 large medium effort, perhaps  around glm 5.3 flash level . Qwen qoduck is apparently qw4 27b, from the limited outputs, it seems to be  better than q3.8 flash next in some tasks And worse than ds v4.1 flash 1017


```



document fetched 2026-11-12 00:44 · canonical https://reddit.com/r/LocalLLaMA/comments/1xtt3rb/observations_from_seeing_qwen_4_maxs_outputs
