News Summary

OpenAI has announced its latest reasoning models, o3 and o3-mini, as part of its ’12 Days of OpenAI’ event. These models, which succeed the o1 models, are not yet widely available but are open for preview to safety researchers. The o3 model reportedly outperforms earlier versions in coding benchmarks and introduces new research on deliberative alignment, requiring the AI to reason about safety decisions. This development marks a significant advancement in AI reasoning capabilities.