o3-pro

2025-06-10 · Publisher : OpenAI · Domain : Artificial Intelligence

OpenAI launches o3-pro, a version of o3 that thinks longer for more reliable answers, available to ChatGPT Pro and Team subscribers and in the API, alongside an o3 price cut.

OpenAI launches o3-pro, a version of o3 that thinks longer for more reliable answers, available to ChatGPT Pro and Team subscribers and in the API, alongside an o3 price cut. ### What changes - Replaces o1-pro in ChatGPT; same tools as o3 (web search, file analysis, vision, Python, memory). - Expert reviewers prefer it to o3 in every category tested, notably science, education, programming and business. - Evaluated with a "4/4" reliability test: the model must answer correctly in all four attempts. - Responses often take several minutes, so it is best reserved for difficult questions. - The price of o3 in the API is cut by 80% the same day. ### Specifications - Context 200,000 tokens, output 100,000 tokens, text and image input, text output. - Knowledge cutoff: June 1, 2024. - API identifier: o3-pro (snapshot o3-pro-2025-06-10), Responses API only (background mode recommended), Batch supported. - Limits at launch: no image generation or canvas, temporary chats disabled. - ChatGPT: Pro and Team the same day, Enterprise and Edu the following week. ### Pricing Pricing per million tokens (USD), according to the documentation. | | Input | Output | | --- | --- | --- | | o3-pro | 20 | 80 | | o3 after the cut | 2 | 8 | ### Benchmarks Scores published by the vendor: the announcement states that o3-pro outperforms o1-pro and o3 on academic evaluations, with results in charts only. No numerical value is included. ### Key takeaways o3-pro is for cases where reliability matters more than speed and cost. It was replaced by GPT-5 pro from August 2025.

Tags : LLM, OpenAI, o-series, Raisonnement, Pro