%2Fcdn.vox-cdn.com%2Fuploads%2Fchorus_asset%2Ffile%2F25636776%2FVST_0924_Site.jpg)
Performance Breakthroughs
introduced o3 and o3-mini, showcasing significant advancements in reasoning capabilities, surpassing previous models in coding, math, and science benchmarks.
Notably, o3 excelled in the SWE-Bench Verified coding test by 22.8% and achieved a remarkable 96.7% on the AIME 2024 math test, demonstrating its superior problem-solving abilities.
Accessibility and Development
While not yet publicly available, OpenAI invites the research community for early testing of these models, aiming to refine them further before a wider release.
OpenAI emphasizes the ongoing development of o3 and o3-mini, noting that final results may evolve and highlighting the introduction of new safety measures to enhance reliability and reduce deceptive tendencies.