OpenAI Releases O3-Pro: Next-Gen AI Reasoning Model

OpenAI Introduces o3-pro: A New AI Model

OpenAI has recently unveiled o3-pro, positioning it as their most advanced AI model to date.

This new model builds upon OpenAI’s existing o3, a reasoning model initially launched earlier in the year.

How o3-pro Differs from Traditional AI

Unlike many conventional AI systems, reasoning models tackle problems through a sequential, step-by-step process. This approach enhances reliability in complex areas such as physics, mathematics, and software coding.

The structured methodology allows for more dependable results in these specialized domains.

Availability and Rollout

o3-pro is now available to ChatGPT Pro and Team subscribers, effectively replacing the o1-pro model.

Enterprise and Edu users will gain access in the following week, as announced by OpenAI.

Furthermore, o3-pro is currently accessible through OpenAI’s developer API.

Pricing Details

The API pricing for o3-pro is set at $20 per million input tokens and $80 per million output tokens.

Input tokens represent the data provided to the model, while output tokens are the results generated by the model based on that input.

For context, a million input tokens equates to approximately 750,000 words, exceeding the length of the novel “War and Peace.”

Performance and Evaluation

According to OpenAI’s changelog, expert evaluations consistently demonstrate a preference for o3-pro over o3 across all tested categories.

This preference is particularly pronounced in fields like science, education, programming, and writing assistance.

Reviewers have also consistently rated o3-pro higher for its clarity, thoroughness, adherence to instructions, and overall accuracy.

Capabilities of o3-pro

o3-pro is equipped with access to a range of tools, enabling it to perform diverse tasks.

• Web searching
• File analysis
• Visual input reasoning
• Python execution
• Personalized responses through memory
• And more

However, OpenAI notes that responses from o3-pro generally require more processing time compared to o1-pro.

Limitations and Known Issues

Currently, temporary chats with the model within ChatGPT are disabled as OpenAI addresses a technical issue.

o3-pro does not support image generation at this time.

Additionally, OpenAI’s AI-powered workspace feature, Canvas, is not yet compatible with o3-pro.

Benchmark Results

Internal testing by OpenAI reveals that o3-pro achieves impressive results on established AI benchmarks.

On AIME 2024, a test of mathematical abilities, o3-pro outperformed Google’s Gemini 2.5 Pro.

Similarly, o3-pro surpassed Anthropic’s Claude 4 Opus on GPQA Diamond, a benchmark assessing PhD-level scientific knowledge.