# Are there any model architecture innovations/breakthroughs that we're looking forward to in the next wave of open model drops? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/e5416a95b24697a2

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-17 15:39 ingested 2026-09-18 02:42

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

## retained text



```
Basically the title. Training models take time and if a model was already in the process of pre-training/RL then it can't be re-cooked with any new architectural improvement that has been discovered in that timeframe. So, has there been any new innovation which couldn't be incorporated into GLM 5.3, Qwen 3.8, DS V4.1 and such models but we're looking forward to in the next generation? I live under a rock so I'm not aware much of what sort of breakthroughs happen in the industry, but I did hear that the Deepseek V4.1 architecture was really good but couldn't bring out its full potential because it was undertrained.
```



document fetched 2026-09-18 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1wixgpr/are_there_any_model_architecture
