OpenAI has introduced its next-generation reasoning model, o3, marking a significant leap from the earlier o1 model. The o3 series includes both the standard and a mini version, optimized for specific tasks. OpenAI claims that under certain conditions, o3 approaches artificial general intelligence (AGI), capable of outperforming humans in economically valuable tasks. This achievement could alter OpenAI's partnership dynamics with Microsoft, as the agreement stipulates no obligation to share AGI-level technology.
CEO Sam Altman announced plans to release the o3 mini by the end of January, followed by the full version. The o3 model has already set records on the ARC-AGI benchmark, scoring 87.5% in high-computation tests, surpassing human-level benchmarks. It also excelled in other tests, including Codeforces Elo and SWE-bench Verified, indicating superior performance in complex problem-solving compared to previous models.
The competitive AI landscape is heating up, with Google and Meta also advancing their models. OpenAI's o3 debut concludes its 12-day product launch event, showcasing innovations like the new ChatGPT Pro and AI video generator Sora Turbo.
