OpenAI o3-Mini on GPTunneL
GPTunneL keeps expanding its platform, and today we're happy to announce that OpenAI o3-mini is now available. This system delivers high accuracy and speed on tasks in science, technology, engineering, and mathematics (STEM). Below we explain its key advantages.
Advantages of OpenAI o3-mini
The o3-mini model was built with the needs of specialists in mind — people who need fast, accurate answers when solving complex technical problems. Its key advantages include:
- Optimized for STEM tasks. o3-mini handles the most complex reasoning, math, and logic problems, generates correct and optimized code, and provides detailed explanations for all its answers. After each task you can easily review the model's reasoning process and, for example, refine your prompt.
- Flexible reasoning-level configuration. The developers continued improving responsiveness to user prompts. This lets you adapt the model to specific tasks — from quick answers to deep analysis. The best way to see how much OpenAI has improved instruction-following is to compare it with previous versions in our Arena.
- Improved speed and lower latency. Thanks to an optimized architecture, time to first token has been cut by 2500 ms compared to the previous o1-mini version. In practice, this means the model delivers near-instant responses even with large volumes of data.
- A cost-effective alternative without compromises. The cost of using o3-mini is lower than o1, while matching the price of o1-mini. This makes the model accessible to a wide range of users while maintaining high performance and answer quality. You can find the cost of using o3-mini in the "Pricing" section of our website.
Benchmarks and research results
OpenAI's internal testing confirms the high performance of o3-mini. Reports mention three versions: Low, Medium, and High. GPTunneL offers only the Medium version, which shows the following results:
- Math problems (AIME 2024). On competition-level math problems, o3-mini Medium achieved 79.6% accuracy, demonstrating its ability to handle computational tasks. For comparison, its predecessor, o1-mini — until recently considered the best in the field — scores 16% lower.
- PhD-level science (GPQA Diamond). On tests of scientific reasoning, the model showed 76.8% accuracy, highlighting its deep knowledge of biology, chemistry, and physics. For comparison, on the same benchmark o1-mini scores only 63.6%.
- Software engineering (SWE-bench Verified). On software development tasks, the system showed 42.9% accuracy. Here the model slightly outperforms its predecessor, o1-preview.
- Overall quality ratings. OpenAI surveys showed that users preferred o3-mini's answers over o1-mini's in 56% of cases, and the number of critical errors on complex real-world tasks dropped by 39%.
These figures confirm that OpenAI o3-mini not only meets high quality standards but also represents a significant leap forward, outperforming previous versions across several key metrics.
Try OpenAI o3-mini on GPTunneL today!
Don't miss the chance to work with OpenAI's cutting-edge model. Sign up on our platform and experience all the advantages of o3-mini. To use the model without restrictions, just top up your balance with a small amount in your Profile.