ChatGPT / Depositphotos
OpenAI CEO Sam Altman announced that the free version of ChatGPT will feature o3-mini — one of the new “reasoning” models of the company.
At the same time, for owners of paid subscriptions, ChatGPT o3-mini will offer “more options,” so certain limitations are expected in the free version.
Reasoning models are distinguished by the ability to effectively fact-check and avoid some pitfalls that usually “stall” standard versions. They also show more reliable results in natural sciences, physical and mathematical problems.
The o3-mini model is smaller in scale compared to OpenAI’s flagship model GPT-4-turbo, but will offer faster response times (especially for simpler queries) and less computational demands.
Interestingly, the announcement about access to o3-mini in the free version appeared against the backdrop of news about the release of the Chinese DeepSeek R1 — also a reasoning model, but immediately launched for everyone. According to the company, R1 surpasses the previous reasoning model from OpenAI, o1, in AIME, MATH-500, and SWE-bench Verified tests (the first compares the model with others to evaluate effectiveness, the second is a collection of textual problems, and the third focuses on programming tasks).
Earlier, OpenAI had announced plans to release o3-mini by the end of January 2025.