OpenAI unleashes o3-mini reasoning model

OpenAI on Friday released the latest model in its reasoning series, o3-mini, both in ChatGPT and its application programming interface (API). It had been in preview since December 2024.

The company said in its announcement that it “advances the boundaries of what small models can achieve, delivering exceptional STEM capabilities — with particular strength in science, math, and coding — all while maintaining the low cost and reduced latency of OpenAI o1-mini.”

OpenAI said that o3-mini delivered math and factuality responses 24% faster than o1-mini, with medium reasoning effort, and testers preferred its answers to those generated by o1-mini more than half the time.

In addition, the announcement said, “while OpenAI o1 remains our broader general knowledge reasoning model, OpenAI o3-mini provides a specialized alternative for technical domains requiring precision and speed. In ChatGPT, o3-mini uses medium reasoning effort to provide a balanced trade-off between speed and accuracy. All paid users will also have the option of selecting 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 model is now available to users of ChatGPT Plus, Team, and Pro; Enterprise and Education users must wait another week. It will replace o1-mini in the model picker, providing higher rate limits and lower latency. OpenAI is tripling the rate limit for Team and Plus users from 50 messages per day (with o1-mini) to 150 messages per day with o3-mini. The company did not state usage limits for free plan users.

In addition, an early prototype of integration with search will find answers online, with links to their sources.

The model also offers new features for developers who incorporate OpenAI models in their software, including function calling, developer messages, and structured outputs. They can also choose one of three reasoning effort options — low, medium, and high — to adjust power and latency to suit the use case. However, unlike OpenAI o1, it does not support vision capabilities. The company said that o3-mini is available in the Chat Completions API, Assistants API, and Batch API now, to select developers in API usage tiers 3-5⁠.

In addition to its performance, OpenAI touted the model’s safety. “Similar to OpenAI o1, we find that o3-mini significantly surpasses GPT-4o on challenging safety and jailbreak evaluations. Before deployment, we carefully assessed the safety risks of o3-mini using the same approach to preparedness, external red-teaming, and safety evaluations as o1.”