OpenAI Releases Top-Tier Reasoning Model 'o3-pro' - Available via ChatGPT Pro and API

OpenAI officially released its most advanced reasoning AI model, “o3-pro,” on June 10, 2025¹. o3-pro serves as the top-tier model replacing the previous o1-pro model and is now available to ChatGPT Pro and Team users. Enterprise and Edu version users will gain access the following week.

o3-pro is positioned as the flagship version of the o3 family announced in December 2024. Leveraging the characteristics of a reasoning-based model that solves complex problems step-by-step, it has been confirmed to deliver high performance in specialized fields such as physics, mathematics, and programming. According to OpenAI’s internal testing, expert evaluations show that o3-pro outperforms o3 across all test categories, particularly excelling in science, education, programming, business, and writing assistance¹.

Benchmark Performance Surpasses Google and Anthropic

Particularly noteworthy in o3-pro’s performance evaluation is its achievement of results that surpass major competing models. In the AIME 2024 benchmark evaluating mathematical ability, it recorded a score higher than Google’s top-tier model Gemini 2.5 Pro¹. Furthermore, in the GPQA Diamond test evaluating PhD-level scientific knowledge, it demonstrated performance that even surpasses Anthropic’s recently released Claude 4 Opus¹.

The overall performance improvement of the o3 family is also remarkable, with o3-mini’s high-precision mode achieving 96.7% accuracy, a significant improvement from o1’s 83.3%². Particularly in mathematics, o3 recorded a score of 87.7% and demonstrated powerful reasoning capabilities in graduate-level biology, physics, and chemistry problems. Additionally, in the Frontier Math benchmark, while other models remained below 2%, o3 achieved a breakthrough result of 25.2%, according to OpenAI’s announcement².

Diverse Tool Integration and Enhanced Practicality

A major characteristic of o3-pro lies in its multifunctionality that extends beyond mere language processing. The model provides access to a wide range of tools including web search, file analysis, visual input reasoning, Python execution, and personalized responses utilizing memory¹. This enables the processing of more complex and practical tasks.

API pricing is set at $20 per million input tokens and $80 per million output tokens¹, representing a pricing structure designed for enterprise-level utilization. Additionally, o3-mini achieves a 63% cost reduction compared to o1-mini, delivering 24% faster responses with 39% fewer critical errors, demonstrating improved cost performance².

Current Limitations and Future Development

While o3-pro boasts high performance, several limitations exist. The most notable is response speed, with OpenAI reporting that processing takes longer compared to o1-pro¹. Additionally, due to technical issues, temporary chat functionality in ChatGPT is currently disabled, and image generation features and Canvas (OpenAI’s AI-assisted workspace feature) are not supported¹.

OpenAI states that “we recommend using it for challenging questions where reliability is more important than speed and where a few minutes of wait time is a worthwhile tradeoff”¹, positioning o3-pro as a model specialized for professional tasks requiring high precision.

OpenAI’s o3-pro establishes a new benchmark in AI reasoning capabilities, outperforming competitors like Google’s Gemini 2.5 Pro and Anthropic’s Claude 4 Opus in critical areas such as mathematics and science. The model’s breakthrough performance - including 25.2% accuracy on Frontier Math where others achieve below 2% - demonstrates a significant leap in AI’s ability to tackle complex, specialized problems.

While the slower response times and current feature limitations present practical constraints, o3-pro’s strengths in reliability and accuracy make it valuable for high-stakes applications where precision matters more than speed. The integration of diverse tools from web search to Python execution, combined with enterprise-friendly API pricing, positions o3-pro as a powerful solution for organizations requiring advanced reasoning capabilities for research, technical analysis, and complex problem-solving tasks.