OpenAI has launched o3-pro, a refined version of its most advanced artificial intelligence model, offering users more reliable and in-depth responses—particularly for complex tasks. The update is now live for Pro and Team users on ChatGPT and via API access.

Positioned as a successor to o1-pro, the newly released o3-pro builds on the capabilities of OpenAI’s o3 model, with enhanced performance in precision-heavy domains such as science, mathematics, education, and programming.

While it shares the same foundational architecture as o3, OpenAI says o3-pro stands apart in clarity, accuracy, and instruction-following, according to internal expert reviews. The model also surpasses its predecessors in OpenAI’s “4/4 reliability” evaluations, which measure whether a model can consistently provide correct answers across repeated attempts.

Like o3, the o3-pro model is equipped with tool access—allowing it to search the web, analyse files, process visual inputs, run Python code, and recall personalised memories. However, OpenAI cautions that users should expect longer response times due to these added capabilities.

“We recommend using it for challenging questions where reliability matters more than speed, and waiting a few minutes is worth the tradeoff,” the company noted.

Starting today, o3-pro is available in the model picker for Pro and Team subscribers, replacing o1-pro. Enterprise and Edu users will receive access next week. Temporary chats, image generation, and Canvas features are currently unavailable on o3-pro due to ongoing technical adjustments.

Safety information remains unchanged, as o3-pro shares the same underlying model as o3. Those details are outlined in the o3 system card.