Hey @ArtificialAnlys, could you rerun the openai o3 benchmarks and check for metric updates? Lot of people suspecting that the model was quatized.