Script Planner Assistant▸ 
AI Apps 
AI News OpenAI Unveils New Reasoning Models: O3-Mini and O3-Mini-High
小编 ·
OpenAI Unveils New Reasoning Models: O3-Mini and O3-Mini-High
On Friday, during Day 12 of its "12 Days of OpenAI" livestream, OpenAI CEO Sam Altman announced the company's latest AI "reasoning" models, o3 and o3 mini, which are built upon the o1 model released earlier this year.
o3-mini and o3-mini-high focus on logical reasoning, coding, scientific computation, and any tasks requiring deep thinking, logical reasoning, analysis, or creative thinking.
All paid users can also choose "o3-mini-high" in the model selector for a higher-intelligence version, which takes longer to generate responses. Pro users will have unlimited access to o3-mini-high.
Features of o3-mini
Adaptive thinking time: o3-mini allows users to choose the thinking time based on task complexity, offering low, medium, and high reasoning intensity options. This flexibility enables users to find the optimal balance between performance, cost, and latency.
Efficient reasoning capability: Although o3-mini is a smaller model, it performs excellently in handling complex problems and can quickly generate high-quality answers. It has made significant progress in fine-tuning for specific tasks, especially in data analysis and programming tasks.
Economy: o3-mini is designed to deliver efficient performance while reducing computational costs, making it suitable for a wide range of users, including those who need powerful capabilities but have limited budgets.
Features of o3-mini-high
Enhanced computational capability: o3-mini-high outperforms o3-mini in computational capability, making it suitable for tasks requiring higher reasoning intensity. It can handle more complex logical and mathematical problems, making it suitable for professional users and developers.
Higher performance: In high-compute mode, o3-mini-high's performance surpasses that of o3-mini, enabling it to complete more complex tasks in shorter timeframes, suitable for application scenarios with strict response time requirements.
Support for more features: o3-mini-high also supports more complex function calling and structured outputs, enhancing the developer experience and making it more flexible in API integration and application development.
Where is OpenAI o3-mini?
OpenAI o3-mini is now available in ChatGPT and via the API. Pro users will have unlimited access to o3-mini, and Plus & Team users will have triple the rate limits (vs o1-mini). Free users can try o3-mini in ChatGPT by selecting the "Reason" button beneath the message composer.
The launch of o3 and o3-mini marks further advancement for OpenAI in reasoning models. Notably, on the ARC-AGI benchmark, the o3 model reached a level comparable to humans, while o3-mini surpassed the o1 model on certain tasks.
The adaptive thinking time feature of o3-mini allows users to choose different processing speeds based on the needs of specific tasks, providing greater flexibility in practical applications. Additionally, o3-mini's efficiency and economy in handling complex problems make it an ideal choice for users with limited budgets.
o3-mini-high further enhances computational capability and performance, especially in high-compute mode, where it can handle more complex logical and mathematical problems, making it suitable for professional users and developers. It is designed to meet application scenarios with strict response time requirements, supporting more complex function calling and structured outputs, thereby enhancing the developer experience.
Regarding why the model is called "o3" rather than "o2," OpenAI stated that it was to avoid potential trademark conflicts with the UK telecom provider O2.
As other companies develop their own SR models, including Google's announcement of Gemini 2.0 Flash Thinking on Thursday. In November, DeepSeek launched DeepSeek-R1, while Alibaba's Qwen team released QwQ.
