OpenAI says it is starting safety testing and red teaming for o3 and is offering early preview access to safety researchers.
OpenAI states o3-mini supports the same features as o1 (e.g., function calling and structured outputs) while being significantly cheaper and faster in demos.