OpenAI raises the bar with o3-pro, a model that prioritises accuracy over speed
o3-pro builds on the architecture of the base o3, but it’s been fine-tuned to think harder and respond better.
By now, it feels like every other week brings another headline: “OpenAI drops its most powerful model yet,” and that’s not an exaggeration. The company’s rollout schedule has been relentless. After introducing o3 and o4-mini in April, OpenAI teased a pro version on the horizon.
Now, that promise has landed just weeks after with the arrival of its o3-pro, a more powerful version of the o3 model designed to approach problems with precision and accuracy.
Under the hood, o3-pro builds on the architecture of the base o3, but it’s been tuned to think harder and respond better. Pro models like this one are built to take more time answering queries, which means they can think deeper and deliver more precise responses.
Emmanuel Oyedeji
That deliberate style makes it especially strong in math, science, programming, and other areas where accuracy matters more than speed.
But the tradeoff is speed. o3-pro isn’t as fast as its predecessor, o1-pro, but what it lacks in quick replies, it makes up for in thoughtful ones.
And, OpenAI says o3-pro is its most capable model yet, noting that expert reviewers preferred it across the board—for clarity, precision, and instruction-following. In academic benchmarks, it outperformed not only o3 but also major competitors like Google’s Gemini 2.5 Pro and Anthropic’s Claude 4 Opus.
Emmanuel Oyedeji
Even with those strengths, o3-pro launches with a few asterisks. Temporary chats are disabled for now due to a technical issue. It can’t generate images, and it doesn’t support Canvas. OpenAI hasn’t said when those features will roll out to the model.
Meanwhile, the model comes with full access to ChatGPT tools, including Python, web browsing, file analysis, and visual reasoning.
It’s available now for ChatGPT Pro and Team users, replacing o1-pro, with Enterprise and Education rollout coming next week. Through the API, it’s priced at $20 per million input tokens and $80 for output, roughly 750,000 words in, and that same amount out.
The company is betting that reliability, clarity, and domain expertise are what serious users need most. And for developers, researchers, educators, and business professionals, that tradeoff between speed and certainty may be exactly what makes o3-pro the new default.
Emmanuel Oyedeji
Behind the scenes, the company is making bigger moves. It just inked a cloud deal with Google, a surprising move, given that the two are direct rivals in the AI arms race
It is a pivot away from exclusive reliance on Microsoft Azure, which had been its exclusive cloud provider until earlier this year. It also built on a $12 billion agreement with CoreWeave earlier this year to expand its compute capacity.
At the same time, OpenAI is caught in a legal tangle that could reshape how it handles data. The company is appealing a court order that forces it to retain ChatGPT data indefinitely in a copyright case brought by The New York Times. CEO Sam Altman says the ruling threatens user privacy and sets a dangerous precedent.
