Announced for launch in April 2025o3 and o4-mini modelsLater, OpenAI launched the o3-pro model, which it touted as its most powerful artificial intelligence model to date. It will first be available to users of paid subscriptions to ChatGPT Pro and ChatGPT Team services. It will also replace the existing o1-pro model and provide API access resources to developers so that they can create more apps based on the o3-pro model.
o3-pro can be used for web searches, document analysis, inference based on visual input, and Python execution. It also supports contextual connections using a memory function. However, due to its larger parameter size, response time is slightly longer than o1-pro. Furthermore, it currently cannot generate images and cannot be used in the ChatGPT workspace interface, Canvas.
According to test data released by OpenAI, o3-pro performed better than Google Gemini 2.5 Pro in mathematical computing capabilities in mainstream artificial intelligence benchmarks, and performed better than Claude 4 Opus recently released by Anthropic in knowledge content recognition tests.
As for the API access price, o3-pro costs $20 per million word inputs and $80 per million word outputs.
