OpenAI’s o3 claimed 25%, independent test says “try 10”
OpenAI’s o3 AI model scored lower on the FrontierMath benchmark than the company initially implied, according to independent tests by Epoch AI, the research institute behind FrontierMath. When OpenA...