OpenAI has introduced a new AI model — o3-mini, available in ChatGPT and via API. It will provide faster and more accurate answers in science, mathematics and programming.
This is stated in the company's press release.
The o3-mini is the first compact model to include popular developer features such as function calls, structured results, and the ability to receive messages.
The model will support streaming and give developers a choice between three levels of compute resources — low, medium, and high — to optimize performance based on specific needs.
In addition, the o3-mini is faster than its predecessor, the o1-mini, and will therefore offer increased speed limits. In particular, Plus and Team users can now send up to 150 requests per day.
In addition, the new model integrates search functions, allowing you to find relevant answers with links to relevant sources, although this feature is still in the testing phase.
Users can try o3-mini for free, and this is the first time such a model has been available to free users in ChatGPT.
Test results confirmed that o3-mini will provide more accurate and understandable answers compared to o1-mini. Users reported a 39% reduction in key errors when performing complex tasks, and response times were reduced: the o3-mini responded to queries 24% faster, with an average response time of 7.7 seconds.
In terms of security, the o3-mini outperformed its predecessors in security tests, demonstrating reduced risks of unauthorized access.