Altman says internal post-Astra OpenAI model can solve problems the world's best mathematicians cannot

rohanpaul_ai · x · 2026-09-16

In a conversation with Salesforce CEO Marc Benioff, Sam Altman laid out a stepped ranking of OpenAI models' math ability: GPT-5.5 is roughly as good as an average math professor, GPT-5.6 as good as a top 1-2 percentile professor, Astra slightly better than that — and an internal model past Astra "can do things that the best mathematicians in the world cannot."

It's a notably concrete description of internal frontier progress, implying the unreleased internal model has surpassed world-class human mathematicians on certain problems. Full video is on the Salesforce YouTube channel.

Related event: Altman: Internal OpenAI Model Beats Top Mathematicians(2 posts)→

Original post →

More from Companies & People

Companies & People channel →