# Astra 6.1 delay? — retained record

> Retained reddit record from reddit:OpenAI.

Canonical: https://highsignal.app/data/records/573998de9076aa9b

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:OpenAI)  published 2026-09-29 06:31 ingested 2026-09-30 02:40

[original source ↗](https://www.reddit.com/r/OpenAI/comments/1wt2qld/astra_61_delay/)

## retained text



```
Are we at a place where frontier LLMs are hitting the laws of diminishing returns hard, so Open AI / Anthropic / etc are doing more and more crazy stuff with training, which makes their models unstable and less useful? They are only going 'off the rails' because it's in their training / weights or due to very poor testing environments and tasks
```



document fetched 2026-09-30 00:30 · canonical https://reddit.com/r/OpenAI/comments/1wt2qld/astra_61_delay
