OpenAI unveils o3 models claiming advancements towards AGI with new reasoning capabilities
2 days ago
OpenAI announced the o3 model family, successor to o1, during its “shipmas” event. The o3 and o3-mini models are claimed to approach AGI under certain conditions. OpenAI skipped o2 to avoid trademark issues. A preview for o3-mini starts soon, but o3 is not widely available yet. The model uses “deliberative alignment” and can adjust reasoning time for better performance. Internal benchmarks show o3 outperforms o1, but external validation is still needed.