OpenAI Releases o3-pro: Its Most Reliable AI Model Yet

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.

The o3-pro model comes equipped with powerful capabilities, including advanced web browsing, Python code execution, vision-based reasoning, file analysis, and support for memory personalization in ChatGPT. However, the release does come with some temporary limitations. At launch, it does not support image generation, the Canvas workspace, or temporary chat sessions, and its response times are slower than earlier models due to the compute intensity of its deeper reasoning capabilities. Early developer feedback notes that complex responses may take up to two to three minutes.

OpenAI CEO Sam Altman also confirmed that the previously expected mid-June release of an open-weights model has been delayed to later this summer, though the company reaffirmed its commitment to open AI development.

Despite the latency trade-offs, early reviews suggest that o3-pro sets a new bar for consistent, trustworthy AI performance. The model is expected to be particularly valuable in domains requiring accuracy and repeatability, such as legal analysis, scientific research, financial modeling, and enterprise automation. The combination of precision, flexibility, and new pricing signals OpenAI’s intent to lead both the consumer and developer markets into a more reliable and robust era of generative AI.

Getting Started of o3-pro

CometAPI provides a unified REST interface that aggregates hundreds of AI models—under a consistent endpoint, with built-in API-key management, usage quotas, and billing dashboards. Instead of juggling multiple vendor URLs and credentials.

Developers can access  O3 API (model name: o3/ o3-2025-04-16)those Deadline for article publication through CometAPI. To begin, explore the model’s capabilities in the Playground and consult the API guide for detailed instructions. Before accessing, please make sure you have logged in to CometAPI and obtained the API key. CometAPI offer a price far lower than the official price to help you integrate.

Also check out o3-Pro API and wait for it to go online soon