According to OpenAI’s release notes, it scores about 10% better than o3 in select benchmarks.
It is available for Pro and Team users, replacing the o1-pro model. It should be available to Enterprise and Edu users next week.
o3-pro also has access to tools, and can therefore search the web, analyze files, take visual inputs and use Python, OpenAI says.
OpenAI o3-pro has access to tools that make ChatGPT useful—it can search the web, analyze files, reason about visual inputs, use Python, personalize responses using memory, and more.
— OpenAI (@OpenAI) June 10, 2025
Worth the tradeoff
It is not, however, able to generate images and does not support Canvas, and responses «typically take longer than o1-pro to complete.»
OpenAI recommends using it for questions where reliability matters more than speed, and says waiting a few minutes is well worth the tradeoff.
Reviewers in «expert evaluations» prefer o3-pro over o3 in every tested category, OpenAI says — «especially in key domains like science, education, programming, business, and writing help.»
The model is also available in the API, costing $20 for a million tokens input, and $80 for a million output tokens, according to TechCrunch — so it’s relatively expensive.
o3 drops 80%
At the same time, Sam Altman announced a welcoming price drop in the API costs for the base o3 model:
we dropped the price of o3 by 80%!!
excited to see what people will do with it now.
think you'll also be happy with o3-pro pricing for the performance 🙂
— Sam Altman (@sama) June 10, 2025
The cost of o3 is now at $2 per million input tokens, and $8 per million output ones, VentureBeat reports.
That should make the price of what was, until today, OpenAI’s most competent model cheaper than Google’s Gemini 2.5 Pro Preview, which costs about $2.50 for a million input tokens and $15 per million in output, and performs about the same.
Read more: OpenAI’s release notes, a writeup on TechCrunch, Engadget, VentureBeat on the 03 price drop, and teknotum on the launch of o3.