OpenAI has unveiled o3-pro, a new version of its reasoning-focused AI model. The rollout, which began on June 10, is currently available to ChatGPT Pro and Team users, replacing the previous o1-pro. Enterprise and education users are slated to gain access next week. O3-pro builds upon the earlier o3 model, released earlier this year. It’s designed for enhanced reliability, particularly in coding, mathematics, and science. OpenAI indicates that this model excels in extended thinking tasks, making it more suitable for scenarios where precision is paramount. The model is slower than its predecessor, but offers superior accuracy. Internal evaluations at OpenAI showed a preference for o3-pro across various areas, including science, programming, business inquiries, and writing assistance. Benchmarks demonstrate o3-pro’s superiority. On the AIME 2024 mathematics benchmark, o3-pro surpassed Google’s Gemini 2.5 Pro. Additionally, it outperformed Anthropic’s Claude 4 Opus on the GPQA Diamond test, which assesses PhD-level science understanding. The model’s API pricing is set at $20 per million input tokens and $80 per million output tokens. Temporary chats are currently unavailable due to a technical issue, and image generation is also unsupported. The model is accessible directly from the model picker for ChatGPT Pro users, and it is suited for tasks requiring accuracy in long-form research, coding support, and content creation. The same safety measures as those in o3 have been implemented in o3-pro.
