OpenAI o3 and o3-mini models bring advanced reasoning capabilities to AI. Available in early access for safety researchers, these frontier models aim to redefine AI safety evaluations and risk mitigation with cutting-edge innovation.