1 post
GPT-5.4 Pro solved an open math problem. The benchmark rated it 'moderately interesting.' The gap between the headline and the scorecard is where the real story lives.