Leaker: Opus 5.5 beat expectations and sped up GPT-6.1 Astra
A leaker reconciles two accounts: OpenAI knew Opus 5.5 would be strong but was caught off guard by how strong it turned out, which increased the urgency around 6.1 Astra.
Small note on something @synthwavedd posted, because I don’t think our reports are necessarily contradictory. My information is that OpenAI already had a good idea that Opus 5.5 was going to be extremely strong, so the release itself and the fact that the model was excellent were not complete surprises internally. Leo’s information seems to be slightly different: that people were still caught off guard by just HOW strong the final performance turned out to be, and that this increased the urgency around 6.1 Astra. Both things can easily be true at the same time. Knowing a competitor has a very strong model coming is not the same as fully anticipating how impressive the final model will feel in practice once it lands. We’re also receiving information from different people inside different parts of these companies, so small variations like this are completely normal. For now, I still consider my original information valid, and I also think Leo’s report adds useful context rather than directly contradicting it. 😊
This post quotes another post: “Speculative Q&A weighs whether OpenAI will answer Opus 5.5 by DevDay”