OpenAI is stepping up its game with the release of o3 Pro, a powerful upgrade to its recent multimodal model. With this launch, the company responds directly to recent advances from Gemini 2.5 Pro and Claude 4, placing itself once again at the center of the generative AI race.
o3 Pro builds on o3 with human-rated superiority
The new o3 Pro model enhances the architecture of o3 by focusing on deeper reasoning and refined tool integration. According to OpenAI, human reviewers preferred o3 Pro over o3 in tasks such as scientific analysis, personal writing, data interpretation, and programming, with differences reaching up to 66.7 %.
A leap forward in code, science, and maths
In programming benchmarks like Codeforces, o3 Pro reached an impressive 2 748 Elo score, outperforming the original o3 by over 200 points and landing in the top 200 among human programmers. It also shows improvements in key academic benchmarks such as AIME 2024 and GPQA Diamond, although comparisons with Claude and Gemini remain unclear due to differing test versions.
Pricing, availability, and performance trade-offs
Available now to ChatGPT Pro users and developers, o3 Pro comes with significant cost reductions. It’s 87 % cheaper than o1 Pro and 20 % less expensive than GPT-4o. However, it’s also heavier and slower, making it best suited for users prioritizing reliability over speed.
A strategic move in a competitive AI landscape
While Google and Anthropic offer compelling alternatives, OpenAI’s release strategy with o3 Pro suggests a clear shift toward specialised performance and affordability. The absence of third-party benchmarks leaves some open questions, but o3 Pro seems poised to compete at the highest level.