San Francisco, June 10, 2025 — OpenAI has officially launched o3-pro, its most capable and reliable language model to date, now available to ChatGPT Pro and Team users. Replacing the previous o1-pro, the o3-pro model is also rolling out to API developers immediately, with Enterprise and Education accounts gaining access next week. The release emphasizes advanced reasoning, consistent reliability, and improved tool integration, targeting professional and enterprise-grade use cases.
According to OpenAI, o3-pro outperforms both o1-pro and the earlier o3 model in internal and external benchmarks, particularly in science, mathematics, coding, and high-stakes decision-making. The model achieves high scores in OpenAI’s “4/4 reliability” testing—designed to measure whether a model can consistently produce correct answers across repeated queries. Compared to leading competitors such as Google’s Gemini 2.5 Pro and Anthropic’s Claude 3 Opus, o3-pro has shown superior performance on rigorous benchmarks like AIME (math) and GPQA (graduate-level science).
With this release, OpenAI also announced a major shift in pricing strategy. While o3-pro is positioned as a premium model—priced at $20 per million input tokens and $80 per million output tokens—the company has simultaneously reduced the cost of the base o3 model by 80%, making it more accessible to developers and startups at $2 and $8 per million input and output tokens respectively.