# Qwen4 expectations — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/e173706f82c2192f

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-26 09:12 ingested 2026-10-27 02:40

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xfrti9/qwen4_expectations/)

## retained text



```
Not trying to predict which model sizes they're going to release, but seeing how massively impressive Qwen3.8-Flash-Next is for its size despite being a preview model meant for showcasing and implementing the new arch, it feels like when Qwen4 hits I'm not going to be needing larger models at all anymore.

  
Do we expect Qwen4 to exceed the current QFN by a lot, or de we believe the current model is already hitting near the optimum we can expect today for this size?
```



document fetched 2026-10-27 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xfrti9/qwen4_expectations
