OpenAI has officially rolled out its new o3-pro model, offering enhanced reasoning capabilities to all ChatGPT Pro and Team users starting this week. The o3-pro model replaces the previous o1-pro and is also now available through the OpenAI API. Enterprise and educational users are expected to gain access within the next few days, marking a broad deployment across OpenAI’s ecosystem.
The release lands as ChatGPT now handles around 2.5 billion prompts a day.
The o3-pro model builds on the foundational o3 architecture but brings notable improvements in critical areas such as science, programming, data analysis, education, and writing. According to OpenAI, o3-pro is designed to “think longer,” prioritizing reliability and depth over response speed, especially when handling complex or high-stakes queries. In expert evaluations, o3-pro has consistently outperformed the standard o3 model in terms of clarity, comprehensiveness, and accuracy.
See Sam Altman’s announcement on X →
OpenAI has confirmed that o3-pro integrates seamlessly with tools already available in the o3 environment, including web browsing, file analysis, Python execution, visual input interpretation, and personalized responses using memory. These features make the model especially useful for users who rely on AI for tasks that require deeper analysis or step-by-step reasoning.
In performance benchmarks, o3-pro has delivered strong results. It has shown exceptional capabilities on the AIME 2024 assessment, which tests mathematical aptitude, and on the GPQA Diamond benchmark, which measures PhD-level science knowledge. In both areas, it outperformed leading models from competitors like Google and Anthropic, solidifying OpenAI’s position in the competitive landscape of advanced AI systems.
The API pricing for o3-pro has also been announced. Input tokens are priced at $20 per million, while output tokens cost $80 per million. This reflects its status as a high-performance model aimed at users who need precision and depth, such as researchers, analysts, and advanced developers.
There are still a few limitations. As of now, o3-pro does not support image generation or access to the Canvas interface within ChatGPT, and temporary chats are currently disabled due to a technical issue. Additionally, users may notice that responses take slightly longer to generate compared to the earlier o1-pro model due to the model’s more thoughtful reasoning process.
In related news, OpenAI has significantly reduced the price of the standard o3 model by 80%, making it far more affordable for developers who don’t require the advanced capabilities of o3-pro. CEO Sam Altman also revealed that the company is delaying the release of its open-weights model, which was originally planned for June. The delay comes as a result of what Altman described as an “unexpected and quite amazing” development from the research team, though no further details have been shared.
The launch of o3-pro highlights OpenAI’s continued focus on building AI systems that are not just fast, but also trustworthy and well-suited for complex tasks. As the technology landscape becomes more demanding, models like o3-pro are designed to meet the growing need for reliability, nuance, and deeper reasoning across a wide range of professional use cases.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →