OpenAI Unveils O3-PRO: The Next Frontier in AI Reasoning Capabilities

OpenAI has just released its upgraded AI model, O3-PRO, a significant enhancement over its predecessor, O3. This new model is lauded as the most powerful AI reasoning model to date, designed to approach complex problems systematically over a variety of domains, particularly in physics, mathematics, and programming.

Starting this Tuesday, O3-PRO became available to subscribers of ChatGPT Pro and Team, phasing out the older O1-PRO model. OpenAI plans to roll out access to Enterprise and Edu users the following week. Moreover, O3-PRO has also hit OpenAI’s developer API, further expanding its reach.

With a pricing structure of $20 per million input tokens and $80 per million output tokens, O3-PRO is aimed at providing cost-efficient AI solutions. For context, a million input tokens can encompass approximately 750,000 words, which is somewhat longer than a classic novel like “War and Peace.”

Reviewers conducting expert evaluations have consistently noted that O3-PRO surpasses O3 across all tested criteria, achieving especially high praise in critical areas including science, education, programming, business strategy, and writing supports. The model has been rated higher in clarity, comprehensiveness, instruction-following, and above all, accuracy.

Equipped with innovative tools, O3-PRO can search the internet, analyze files, interpret visual inputs, and execute code through Python, marking a broad advance in its functional capabilities. However, it’s important to note that responses from O3-PRO may take longer to generate compared to those from O1-PRO.

Some inherent limitations also exist; users currently cannot engage in temporary chats with O3-PRO in ChatGPT due to ongoing technical issues. Additionally, the model lacks image generation capabilities, and OpenAI’s AI-powered workspace feature, Canvas, is currently unsupported.

Despite these caveats, O3-PRO has proudly achieved remarkable results in leading AI benchmarks. For instance, in the AIME 2024 assessments, it outperformed Google’s elite AI, Gemini 2.5 Pro, in mathematical prowess. It also exhibited superior results to Anthropic’s Claude 4 Opus on the GPQA Diamond test, which measures PhD-level science knowledge.

As AI continues to evolve, O3-PRO marks a pivotal moment in the landscape of reasoning technology, cementing OpenAI’s position at the forefront of AI advancements.