OpenAI Launches o1 Reasoning Model, Touting Major Gains in Math, Coding and Science
Read the full article for context, quotes, and updates from the team.
OpenAI has unveiled o1, a new family of models built to tackle complex reasoning tasks in math, coding and science. The company says the system is designed to spend more time “thinking” before responding, using reinforcement learning to improve step-by-step problem solving rather than relying only on fast pattern matching.
According to OpenAI, o1 delivers strong results on difficult benchmarks. The model scored 83% on AIME 2024, a challenging mathematics test, and 74% on GPQA Diamond, a benchmark focused on graduate-level science questions. OpenAI says these results mark a significant improvement over earlier models, including GPT-4o, especially on tasks that require multi-stage reasoning and careful analysis.
The new model family is intended for users who need more advanced performance on technical and analytical work, including researchers, developers and students. OpenAI said o1 is now available to ChatGPT Plus and Team users, with wider access planned in the near future.
The launch highlights the company’s push to improve AI systems that can handle harder problems more reliably, a key area of competition across the industry as developers race to build models with stronger reasoning capabilities.
