AI _ Post_OpenAI new "O1" model - Post

OpenAI has unveiled "o1," its first "reasoning" AI model, aimed at solving complex questions faster than humans. Alongside it, OpenAI released a smaller, more affordable version called "o1-mini." The o1 model demonstrates superior performance in tasks such as coding and multistep problem-solving compared to previous models, although it is slower and more expensive to use than GPT-4o. The training approach for o1 differs, utilizing a new optimization algorithm and reinforcement learning, allowing it to solve problems independently.

OpenAI emphasizes that while o1 hallucinates less than previous models, the issue hasn't been entirely resolved. The model excels in mathematical challenges, performing significantly better on tests like the AP math exam and the International Mathematics Olympiad than its predecessors. However, it lacks GPT-4o's broader knowledge of factual data and doesn't have web browsing or file processing capabilities.

The model has been designed to show its reasoning in a way that mimics human thought, although OpenAI acknowledges it doesn't actually think like a human. Access to the model is available to ChatGPT Plus and Team users, with broader access planned, though developer access comes at a high cost.