o3-pro Now Available to ChatGPT Pro Users, Open-Weights Model Coming Later

OpenAI has officially announced the rollout of its advanced language model, o3-pro, which is now being gradually made accessible to users. As part of this release, ChatGPT Pro and Team subscribers will notice o3-pro appearing in their model selection options, replacing the earlier o1-pro model. Meanwhile, Enterprise and Education clients are expected to gain access to the new model starting one week after this initial deployment.

According to OpenAI, the o3-pro model offers significantly improved capabilities across a variety of domains. These enhancements were validated through expert assessments, where o3-pro was consistently rated higher than its predecessor, o3. The model demonstrated superior aptitude in fields such as education, scientific research, computer programming, technical writing, and data interpretation. Reviewers also reported a noticeable boost in the model’s ability to follow instructions, provide comprehensive responses, and generate content with greater clarity and precision.

Much like its predecessor, o3-pro supports an array of advanced features. These include the ability to browse the internet for real-time information, interpret visual content, analyze uploaded documents, and conduct data-driven tasks using Python. Additionally, it retains the memory function to personalize user interactions, making conversations more relevant and adaptive over time.

On June 11, 2025, OpenAI CEO Sam Altman posted on X (previously Twitter) to confirm the rollout. He expressed excitement over the model’s performance, stating he was initially skeptical about the significant win rates o3-pro demonstrated over o3. Altman referred to the model as “really smart,” underlining its impressive advancements.

However, anticipation surrounding the release of OpenAI’s open-weights model has been met with a slight delay. Altman clarified that while the team had originally aimed for a June release, the model will now be postponed until later in the summer. He cited a breakthrough by the research team as the reason for this delay, noting that the wait will be well worth it due to unexpected and remarkable developments. Unlike a fully open-source model, this open-weights version will provide greater customization and adaptability for developers while marking OpenAI’s first such release since GPT-2 in 2019.