OpenAI has released o3-pro, a refined AI model focused on reasoning. This model, which replaced o1-pro, is available to ChatGPT Pro and Team users from June 10, with enterprise and education users to follow next week. o3-pro is based on the earlier o3 model, designed to be more reliable, especially in coding, math, and science. OpenAI emphasizes its suitability for long-form thinking tasks, prioritizing accuracy over speed. While slower than its predecessor, o3-pro delivers superior outcomes, as confirmed by internal testing across science, programming, business, and writing applications. Benchmarks confirm o3-pro’s leading performance. It exceeded Google’s Gemini 2.5 Pro on the AIME 2024 math benchmark and Anthropic’s Claude 4 Opus on the GPQA Diamond test, which evaluates advanced science understanding. API pricing is set at $20 per million input tokens and $80 per million output tokens. Notably, temporary chats are unavailable due to a technical issue, and image generation is not supported. Accessible through the model picker for ChatGPT Pro users, o3-pro is recommended for detailed research, coding support, and content creation, where accuracy is a priority. o3-pro integrates the same safety measures as o3.
o3-pro: OpenAI’s New AI Model Claims Victory Over Competitors in Science and Math Assessments
