OpenAI has introduced o3-mini, a compact reasoning model offering cost-effective performance.

Available through ChatGPT and API, it improves speed and affordability compared to its predecessor, o1-mini.

The models include features like function calling and structured outputs, with three levels of reasoning effort—low, medium, and high—to adjust settings based on task complexity.

The o3-mini provides better accuracy in technical domains, including AIME math and GPQA science tests.

It also excels in coding and software engineering tasks, especially at higher reasoning levels. The model has a 24% faster response time and enhances safety with improved training techniques

However, o3-mini does not support visual reasoning, which is still handled by the o1 version.

Access is available for ChatGPT Plus, Team, and Pro users, with Enterprise access expected later in February.

Recent OpenAI developments