AI math's pay-to-play era: open-source models can't keep up
A new research-level math benchmark called First Proof shows frontier AI models excelling while open-source rivals struggle, raising cost and access questions for builders.
By Sara Al-Ghamdi·· 3 min




