The release aims to provide developers with a specialised tool that balances speed and accuracy, offering improved performance for complex challenges within the STEM arena.

In the company blog, OpenAI described o3-mini as its first small reasoning model supporting highly requested developer features, including function calls, Structured Outputs, and developer messages, making it production-ready out of the gate.

The model supports streaming and allows developers to choose between three reasoning effort options—low, medium, and high—to optimise for specific use cases.

"This flexibility allows o3-mini to ‘think harder’ when tackling complex challenges or prioritise speed when latency is a concern," OpenAI noted.

While o3-mini does not support vision capabilities, developers are advised to continue using OpenAI o1 for visual reasoning tasks. It is rolling out in the Chat Completions API, Assistants API, and Batch API starting immediately for select developers in API usage tiers 3-5.

OpenAI emphasises that its o1 models remain the flagship reasoning models, but the o3-mini offers a specialised experience for those who need enhanced STEM reasoning.

In ChatGPT, o3-mini utilises medium reasoning effort to provide a balanced trade-off between speed and accuracy. All paid users can select o3-mini-high in the model picker for a higher-intelligence version that takes a little longer to generate responses. Pro users will have unlimited access to both o3-mini and o3-mini-high.

The o3-mini model outperforms o1 in certain situations, especially within the STEM domain. OpenAI said:

"Similar to its OpenAI o1 predecessor, OpenAI o3-mini has been optimised for STEM reasoning. o3-mini with medium reasoning effort matches o1’s math, coding, and science performance while delivering faster responses. Evaluations by expert testers showed that o3-mini produces more accurate and clearer answers, with stronger reasoning abilities, than OpenAI o1-mini. Testers preferred o3-mini’s responses to o1-mini 56 per cent of the time and observed a 39 per cent reduction in major errors on difficult real-world questions. With medium reasoning effort, o3-mini matches the performance of o1 on some of the most challenging reasoning and intelligence evaluations including AIME and GPQA."

OpenAI also highlights the speed and efficiency of the o3-mini model: "With intelligence comparable to OpenAI o1, OpenAI o3-mini delivers faster performance and improved efficiency. Beyond the STEM evaluations highlighted above, o3-mini demonstrates superior results in additional math and factuality evaluations with medium reasoning effort. In A/B testing, o3-mini delivered responses 24 per cent faster than o1-mini, with an average response time of 7.7 seconds compared to 10.16 seconds."