OpenAI has launched o3-pro, an enhanced version of its reasoning model, o3, originally introduced earlier this year. Unlike traditional AI models, reasoning models like o3-pro solve problems step by step, which improves reliability in complex domains such as physics, mathematics, and programming.
Availability and Pricing
Starting Tuesday, o3-pro is accessible to ChatGPT Pro and Team users, replacing the previous o1-pro model. Enterprise and education users will gain access the following week. Additionally, o3-pro is now live in OpenAI’s developer API. Pricing for the API is set at $20 per million input tokens and $80 per million output tokens, where input tokens are the data fed into the model and output tokens are the generated responses. For scale, one million input tokens approximate 750,000 words—slightly longer than War and Peace.
According to OpenAI, expert reviewers consistently prefer o3-pro over the original o3 across multiple categories, particularly in science, education, programming, business, and writing assistance. They rated o3-pro higher for clarity, thoroughness, adherence to instructions, and accuracy.
O3-pro supports an array of tools, enabling it to search the web, analyze files, process visual inputs, execute Python code, and personalize responses through memory features. However, its responses generally take longer to complete compared to o1-pro.
Limitations and Current Restrictions
Some limitations apply: temporary chats with o3-pro in ChatGPT are currently disabled due to a technical issue. The model does not support image generation or OpenAI’s Canvas workspace feature at this time.
OpenAI reports that o3-pro outperforms other leading AI models on respected benchmarks. For instance, it scored higher than Google’s Gemini 2.5 Pro on AIME 2024, a test of advanced math skills. It also surpassed Anthropic’s Claude 4 Opus on GPQA Diamond, which assesses PhD-level science knowledge.
What The Author Thinks
While o3-pro’s benchmarks and expert reviews highlight impressive advancements, the true measure of its impact will be how effectively it performs in real-world scenarios. Longer response times and current feature limitations suggest there’s still work to do before it can fully replace or augment human experts across all fields. Nonetheless, its stepwise reasoning approach marks a promising direction in AI development.
Featured image credit: Tom’s Hardware
For more stories like it, click the +Follow button at the top of this page to follow us.