Comparing Math Proofs: GPT-4o vs. Astra

danshipper · x · 2026-08-03

A developer shared the analysis generated by GPT-4o attempting a mathematical conjecture and compared it directly against the results from OpenAI's frontier model, Astra. This provides further insight into the reasoning capabilities of weaker models versus frontier models.

Related event: Measuring AI Research Ability via "Prompt Distance": Weak Models Reproduce Frontier Math Proofs(6 posts)→

Original post →

More from Models

Models channel →