As promised last week, OpenAI has officially rolled out its latest AI model, o3-mini, to all ChatGPT users, including those on the free plan. This new release brings significant improvements in reasoning, particularly in mathematics, programming, and scientific analysis.
According to OpenAI, the o3-mini model pushes the limits of what compact AI models can accomplish, delivering responses 24% faster than its predecessor, o1-mini. Independent testing indicates that users prefer o3-mini’s answers 56% of the time compared to o1-mini, while also noting 39% fewer errors in its responses.
Like its predecessor, o3-mini displays its reasoning process above responses, allowing users to review its logical steps. Additionally, it can integrate with web searches to enhance accuracy—though this functionality is still in the early development phase.
This launch follows a highly competitive week in the AI landscape, particularly with DeepSeek’s free AI models from China rapidly gaining traction for their impressive speed and accuracy. OpenAI’s response aims to regain user interest and reinforce its dominance in the AI space.
Access and Availability
For free-tier users, o3-mini can be accessed by clicking the “Reason” button in the text input field. While OpenAI has not detailed exact usage limits, it is expected to align with existing restrictions on GPT-4o—allowing a limited number of queries per hour.
For paying users, o3-mini is available via the model selection menu in the top-left corner of the ChatGPT interface. Subscribers to the Plus and Team plans receive 150 queries per day, while Pro plan users enjoy unlimited access for a monthly fee of $200 (approximately £160 / AU$320).
Additionally, premium users gain access to o3-mini-high, a variation of the model that applies deeper reasoning techniques at the cost of slightly longer response times. This enhanced version offers even more accurate and detailed outputs for those willing to wait an extra few seconds.
Safety and Security Enhancements
OpenAI has emphasized the rigorous safety testing behind o3-mini, claiming that it outperforms GPT-4o in detecting unsafe requests and jailbreak attempts. This marks a significant step toward making AI interactions more secure and reliable for users at all levels.
With o3-mini now live, OpenAI continues to refine and expand its AI offerings, ensuring that users—whether free or paid—benefit from more efficient, accurate, and safe AI-powered reasoning
