Photo credit: www.gsmarena.com
OpenAI Launches o3-mini: A New Era in Reasoning Models
In December 2024, OpenAI unveiled the o3-mini, heralded as the company’s “most cost-efficient model” tailored for its reasoning series. This latest addition is particularly adept in fields like science, mathematics, and coding, making it a powerful tool for STEM applications.
OpenAI has indicated that the o3-mini offers a targeted alternative for technology sectors demanding both speed and accuracy. With moderate reasoning capabilities, its performance aligns with that of the o1 model in critical areas such as science, math, and coding, yet it excels in response times. During A/B testing, o3-mini demonstrated a 24% increase in speed compared to o1-mini, averaging a response time of 7.7 seconds, markedly quicker than o1-mini’s 10.16 seconds.
As OpenAI’s inaugural small reasoning model, the o3-mini comes equipped with several much-requested features aimed at developers, which include function calling, developer messages, and Structured Outputs. Furthermore, it will introduce streaming capabilities, allowing developers to select from three levels of reasoning effort—low, medium, and high. The model will also integrate search functionalities, enabling it to provide real-time answers with links to relevant sources accessed online.
Currently, the o3-mini is accessible to subscribers of ChatGPT Plus, Team, and Pro plans. Users can easily select this model from the picker, where it replaces o1-mini. For those willing to exchange response time for enhanced intelligence, the option to choose “o3-mini-high” is also available in the model picker.
Subscribers to ChatGPT Pro will benefit from unlimited access to both the o3-mini and o3-mini-high models, while Plus and Team subscribers will experience an increase in usage limit—from 50 messages per day with the o1-mini to 150 messages when using the o3-mini.
Notably, users of the free ChatGPT version can also experiment with the o3-mini by selecting the “Reason” option in the message composer or by generating new responses, marking it as the first reasoning model provided to non-paying users. The o3-mini is additionally available through Microsoft’s Azure OpenAI Service, broadening its accessibility across various platforms.
Source
www.gsmarena.com