OpenAI has launched o3-pro, an artificial intelligence model that the company says is its most powerful to date. It’s a version of OpenAI’s o3, a reasoning model that the startup launched in April. Meanwhile, Sam Altman announced yesterday that the release of OpenAI’s first open model in years will be delayed until later this summer.
OpenAI has launched o3-pro, an artificial intelligence model that the company says is its most powerful to date. It’s a version of OpenAI’s o3, a reasoning model that the startup launched in April. Meanwhile, Sam Altman announced yesterday that the release of OpenAI’s first open model in years will be delayed until later this summer.
O3-pro is available to ChatGPT Pro and Team users starting Tuesday, replacing the o1-pro model. Enterprise and Edu users will get access in a week, OpenAI said. O3-pro is also available in OpenAI’s developer API, TechCrunch reports .
The price of o3-pro is $20 per million input tokens and $80 per million output tokens in the API.
“In peer-reviewed evaluations, reviewers consistently preferred o3-pro over o3 in every category tested, especially in key areas such as science, education, programming, business, and writing assistance,” OpenAI reports. “Reviewers also consistently rated o3-pro higher for clarity, completeness, adherence to instructions, and accuracy.”
According to OpenAI, o3-pro has access to tools that allow the model to search the web, analyze files, reason about visual data, use Python, personalize its responses using memory, and more. The downside is that the model's responses tend to take longer than those of o1-pro, according to OpenAI.
O3-pro has other limitations. Temporary chats with the model in ChatGPT are currently disabled while OpenAI resolves a “technical issue.” O3-pro cannot generate images. And Canvas, an AI-based feature of OpenAI’s workspace, is not supported by o3-pro.
On the other hand, o3-pro scores impressively on popular AI tests, according to OpenAI’s internal testing. On the AIME 2024 test, which measures a model’s mathematical skills, o3-pro outperformed Google’s best-performing AI model, Gemini 2.5 Pro. O3-pro also outperformed Anthropic’s recently released Claude 4 Opus on GPQA Diamond, a PhD-level test of science knowledge.
OpenAI CEO Sam Altman announced yesterday at X that the open model will be released sometime after June.
“We are going to spend a little more time on our open-source model, meaning we expect it later this summer, but not in June,” he wrote. “Our research team has done something unexpected and quite amazing, and we think it will be very, very worth the wait, but it will take a little more time.”
Recall that Sam Altman announced an 80% reduction in the subscription cost of the most powerful AI model OpenAI o3 .
