The AI models o3 mini and o3 mini high are now available in both ChatPGT and via the API! – Image: Xpert.Digital
Simply smarter: o3-mini and o3-mini-high now for ChatPGT and API applications
OpenAI o3-mini: A quantum leap in cost-effective AI development for sophisticated thinking
The world of artificial intelligence has reached another milestone with the release of OpenAI o3-mini. This state-of-the-art model, now available in ChatGPT and via API, represents the latest and most cost-effective advancement in OpenAI's reasoning model family. Introduced in December 2024, o3-mini redefines the capabilities of smaller models, demonstrating exceptional performance in science, technology, engineering, and mathematics (STEM). It stands out not only for its impressive performance but also for its low cost and reduced latency compared to its predecessor, OpenAI o1-mini.
OpenAI o3-mini marks a paradigm shift by being the first small reasoning model to natively support long-awaited developer features such as function calling, structured output, and developer messages. This means it was designed from the ground up for immediate production use. Like its predecessors o1-mini and o1-preview, o3-mini also supports streaming, enabling flexible and dynamic interaction. Developers also gain the ability to choose between three levels of reasoning effort—low, medium, and high. This flexibility allows the model to be tailored to different use cases: for complex challenges, the reasoning process can be intensified, while for time-critical tasks, the focus is on speed and low latency.
Important note: o3-mini is not suitable for tasks involving visual analysis. For such applications, developers should continue to use OpenAI o1. The rollout of o3-mini will take place in the Chat Completions API, Assistants API, and Batch API, with access initially restricted to selected API users at levels 3-5.
Related to this:
ChatGPT users benefit from enhanced features
There's also good news for end users: ChatGPT Plus, Team, and Pro users can now use o3-mini, while Enterprise customers will gain access in February. o3-mini replaces o1-mini in the model selection, offering higher rate limits and lower latency. This makes it the ideal choice for tasks in programming, STEM, and logical problem-solving. As part of this upgrade, the message limit for Plus and Team users triples from 50 messages per day (with o1-mini) to 150 messages per day (with o3-mini).
Furthermore, o3-mini is now combined with a search function to provide up-to-date answers with relevant sources. This represents an early prototype while OpenAI works to integrate the search function into all reasoning models. This integration allows the model to access a broader range of information and thus generate more informed and up-to-date answers.
Free for everyone to use
A particularly noteworthy new feature is the availability of o3-mini for users of the free ChatGPT plan. They can try out the model by selecting "Reason" in the message composer or by regenerating a response. This is the first time a reasoning model has been offered to free users in ChatGPT and demonstrates OpenAI's commitment to making high-quality AI accessible to everyone.
Starting today (January 31, 2025), users of the free plan OpenAI o3-mini can also try it out by selecting "Reason" in the message composer or by regenerating a response. This is the first time a reasoning model has been made available to free users in ChatGPT.
While OpenAI o1 remains the comprehensive model for general knowledge, o3-mini positions itself as a specialized alternative for technical fields requiring precision and speed. In ChatGPT, o3-mini uses a medium processing power by default to achieve a balanced combination of speed and accuracy. Paying users also have the option to select o3-mini-high in the model selection to use an even more intelligent version that requires slightly more time for its responses. Pro users enjoy unrestricted access to both o3-mini and o3-mini-high.
Performance and optimization for STEM tasks
Like its predecessor o1, o3-mini was specifically optimized for STEM tasks. With medium reasoning effort, o3-mini achieves the same performance as o1 in mathematics, programming, and science, but delivers significantly faster answers. Expert reviews have shown that o3-mini not only provides more accurate and clearer answers, but also demonstrates stronger reasoning ability than o1-mini. Testers preferred o3-mini's answers to those of o1-mini in 56% of cases and observed an impressive 39% reduction in serious errors on challenging, real-world questions. With medium reasoning effort, o3-mini achieves the same performance as o1 on some of the most demanding assessments of reasoning ability and intelligence, including AIME and GPQA. This underscores the model's ability to handle even complex problems.
Detailed performance data in various areas
Competitive mathematics (AIME 2024): In mathematics, o3-mini achieves comparable performance to o1-mini with low cognitive effort. At medium effort, its performance is already on par with o1, while at high effort it surpasses both o1-mini and o1. These results underscore o3-mini's ability to solve mathematical problems with increasing cognitive effort.
PhD-level questions (GPQA Diamond): In the sciences, particularly in biology, chemistry, and physics at the PhD level, o3-mini outperforms o1-mini with low cognitive effort. With high effort, it even reaches the level of o1. This demonstrates that o3-mini is capable of answering and understanding even challenging scientific questions.
Mathematics Research (FrontierMath): In mathematical research, o3-mini outperforms its predecessor in FrontierMath with high computational effort. When using Python tools, o3-mini solves over 32% of problems on the first attempt, including over 28% of the challenging T3 problems. This is clear evidence of the model's ability to solve complex mathematical problems and develop innovative solutions.
Competitive Programming (Codeforces): In competitive programming at the Codeforces level, o3-mini achieves continuously higher Elo scores with increasing computational effort. Even at medium effort, its performance is comparable to o1, while at high effort it significantly surpasses both o1 and o1-mini. These results demonstrate the model's impressive ability to understand and generate code.
Software Engineering (SWE-bench Verified): In the area of software development, o3-mini is the most powerful model that OpenAI has released for SWE-bench Verified to date. This underscores the model's performance in practical software development scenarios.
General knowledge: Data analysis shows clear advantages of o3-mini over o1-mini in various areas of general knowledge. This suggests that o3-mini has developed a more comprehensive understanding of the world.
Speed and efficiency
In addition to its high performance, the o3-mini also stands out for its impressive speed and efficiency. In an A/B test series, the o3-mini was 24% faster than the o1-mini. The average response time was 7.7 seconds compared to 10.16 seconds for the o1-mini. This significant improvement in response times makes the o3-mini an attractive option for applications where speed is critical.
Related to this:
Safety and alignment
OpenAI places great emphasis on the security of its models. To ensure that o3-mini generates secure responses, deliberative alignment techniques were employed. Security assessments show that o3-mini performs better than GPT-4o in tests for challenging security and jailbreak scenarios. This underscores OpenAI's commitment to developing secure and responsible AI systems.
o3-mini: Advanced AI for a smarter future
o3-mini underscores OpenAI's mission to develop cost-effective AI and further democratize high-quality intelligence. With its specialization in STEM tasks, low cost, and remarkable efficiency, o3-mini sets new standards and paves the way for future innovations in artificial intelligence. The model is a vivid example of how AI is becoming increasingly powerful and accessible, with the potential to improve our lives in many ways. The release of o3-mini, as is often said, marks a significant step toward a future where AI is an integral part of our everyday lives.
The development of models like o3-mini is an ongoing process. Continuous improvement of the algorithms, optimization of neural networks, and the collection of ever-larger datasets will lead to even more powerful and efficient AI systems in the future. These advances promise not only to further drive technological progress but also to have a positive impact on many areas of our lives. From medicine and education to the environment – the possibilities are virtually endless.
The “launch of o3-mini,” according to a senior developer at OpenAI, “is another step on the path to solving humanity’s most complex problems with the help of artificial intelligence.” It is this vision that drives OpenAI’s work and leads us toward a future where AI is a powerful tool for improving our world. “We see the development of o3-mini as an important building block that allows us to make intelligence accessible to a wider audience without compromising its capabilities,” he adds.
In summary, OpenAI o3-mini is not only a technological breakthrough but also proof of how advanced technology can be used to change the world and shape a future where artificial intelligence is an integral part of our lives. The “continual development of such models” is, as one AI expert puts it, “a testament to the human spirit of innovation and the relentless pursuit of knowledge and progress.”.
Related to this:
We are here for you - Consulting - Planning - Implementation - Project Management
☑️ Our business language is English or German
☑️ NEW: Correspondence in your native language!
Konrad Wolfenstein
I and my team are happy to be available to you as your personal advisor.
You can contact me by filling out the contact form here wolfenstein@xpert.digital:or simply call me at +49 7348 4088 965. My email address is
I'm looking forward to our joint project.