The subsequent step for OpenAI’s reasoning fashions is o3, a mannequin launched on Dec. 20. o3 was previewed and its smaller cousin, o3-mini, outperformed o1 in coding, math, science and ‘conceptual reasoning’ assessments...