OpenAI Launches o1 Pro, a New Reasoning Model Built for Hard Problems
Read the full article for context, quotes, and updates from the team.
OpenAI has released o1 Pro, a new reasoning model designed to tackle complex tasks in math, coding, and science with greater accuracy than earlier versions. The company says the model outperforms both o1 and GPT-4o on a range of difficult benchmarks, highlighting improvements in multi-step problem-solving and chain-of-thought reasoning.
According to OpenAI, o1 Pro delivers especially strong results on challenging evaluations such as AIME 2024, where it scored 93%, and GPQA Diamond, where it reached 79%. These results suggest the model is better equipped to handle questions that require careful analysis, logical progression, and deeper reasoning rather than simple pattern matching.
The new model is now available to Pro and Team users through ChatGPT and the API. OpenAI positioned the release as a major step forward in AI systems that can support advanced technical work and more demanding analytical tasks.
With o1 Pro, OpenAI is continuing its push toward models that can reason more effectively across complex domains. The launch underscores growing competition in the AI industry to build systems that not only generate fluent responses, but also solve difficult problems with greater reliability.
