OpenAI has announced the release of a new series of AI models designed to enhance reasoning capabilities in ChatGPT and through its API. These advanced models, termed OpenAI o1, are engineered to engage in complex tasks with improved accuracy and problem-solving skills, particularly in fields such as science, coding, and mathematics. Initial tests indicate that the new models outperform previous versions, exemplified by the o1 model's significant improvement in problem-solving accuracy compared to GPT-4o, scoring 83% on challenging mathematics tasks. This development reflects OpenAI's commitment to advancing AI technology while emphasizing safety and adherence to guidelines.
Alongside the release of the o1 series, OpenAI is introducing OpenAI o1-mini, a more compact model that provides faster, cost-effective solutions for coding tasks. Priced at 80% less than the o1-preview, the o1-mini is optimized for developers needing efficient reasoning capabilities without the broader world knowledge of larger models. Both the o1-preview and o1-mini will be accessible to ChatGPT Plus and Team users, with specific message limits. OpenAI is also collaborating with the US and UK AI Safety Institutes to ensure rigorous testing and evaluations of these models, underscoring their commitment to responsible AI deployment.
