o3-pro-2025-06-10

o3-pro is a high-end reasoning model that emphasizes extreme reliability, designed primarily as a replacement for o1-pro, and was officially released by OpenAI on June 11, 2025. It features a context window of 200K tokens and a maximum output token limit of 100K. In the GPQA Diamond scientific test at the doctoral level, o3-pro outperformed Claude 4 Opus; in the AIME 2024 math benchmark, it surpassed Gemini 2.5 Pro. Additionally, the pricing is set at $20 per million input tokens and $80 per million output tokens, representing an 87% price reduction compared to o1-pro.

o3-pro supports long-context analysis and can invoke tools like web search and Python. It also boasts multilingual capabilities, enabling structured JSON output via function calls. Furthermore, in expert evaluations, o3-pro demonstrated superior performance over the o3 model, particularly in areas such as clarity of expression.

Currently, the temporary chat feature for o3-pro has been disabled as OpenAI addresses a technical issue. Additionally, o3-pro does not support image generation—users must rely on GPT-4o, OpenAI o3, or OpenAI o4-mini for image creation—and it also lacks Canvas functionality.