As planned, OpenAI has introduced a new generative AI model called o3-mini, which is now available in ChatGPT and via API. This is the first time that free ChatGPT users can try out models with “reasoning” capabilities alongside paid users. O3-mini operates faster and more efficiently than previous models, while maintaining the ability to solve complex tasks in mathematics, programming, and science.
ChatGPT Pro users will get unlimited access to o3-mini, while Plus and Team plan subscribers will have increased request limits. The new model is also available via the OpenAI API for selected developers, although image analysis is not supported yet.
O3-mini stands out for its accuracy in complex real-world tasks and fewer errors compared to previous models. In particular, it makes 39% fewer serious mistakes on “hard” questions than o1-mini. The model delivers a 24% increase in response speed, making it attractive for developers and users seeking the optimal balance of speed and accuracy.
DeepSeek, a Chinese competitor to OpenAI, recently impressed the market with its R1 product, causing a significant stir in the US. In response, OpenAI accelerated the release of new products in an effort to maintain its market position.
