OpenAI o3 震撼发布！Arc AGI 测试得分超越人类 ｜ OpenAI 12天「第12天」| 回到Axton

OpenAI has introduced O3 and O3 Mini, cutting-edge reasoning models that have outperformed human average scores on various benchmarks, including Arc AGI. Notably, O3 surpassed human performance on the Arc AGI benchmark, marking a significant achievement in AI advancement. While these models are not yet available to the public, OpenAI is launching safety testing for researchers to assess and refine them ahead of their general release, expected around late January for O3 Mini and shortly thereafter for O3. The models have shown remarkable capabilities in coding and mathematics, achieving scores that even surpass some of OpenAI's own researchers in certain areas.