# Qwen 3.8 flash next is based on Qwen 4 architecture, if the announced Qwen 4 27b is also the same architecture with n-grams does it mean I can actually have faster inference on a single 3090 without tweaking much? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/d504a4cecf1c07e4

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-26 19:54 ingested 2026-09-27 02:40

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wqzx12/qwen_38_flash_next_is_based_on_qwen_4/)

## retained text



```
I wish Qwen also released dataset and method to fully train a model ourselves but it is what it is. However, I come here with my stupid question because someone can answer it better. 

And will the model still be an over thinker of faster inference will make up for that. 
```



document fetched 2026-09-27 00:38 · canonical https://reddit.com/r/LocalLLaMA/comments/1wqzx12/qwen_38_flash_next_is_based_on_qwen_4
