OpenAI Introduces New o3 and o3 Mini Models

OpenAI has officially announced its latest artificial intelligence models, named o3 and o3 mini, at the conclusion of its recent 12-day “shipmas” event.

The o3 model serves as the successor to the o1 reasoning model, which was launched in September. Alongside the main o3 model, a smaller version called the o3 mini will also be introduced for specific tasks.

According to OpenAI CEO Sam Altman, the o3 mini model is anticipated to be released by the end of January, with the o3 model following closely thereafter. Prior to the public release, OpenAI has invited safety and security researchers to conduct tests on the new models. Researchers can begin testing the o3 mini model immediately, while the preview for the o3 model will be available at a later date.

Interestingly, OpenAI has decided to skip the naming of the o2 model, moving directly from o1 to o3. A spokesperson for the company explained that this decision was made “out of respect” for a British telecom provider which shares the o2 name.

The new o3 model is reported to have significantly outperformed its predecessors, achieving a 22.8 percent improvement in coding tests. Both the o3 and o3 mini models are designed to be more powerful than the previous o1 models.

In an effort to lead the competitive artificial intelligence landscape, OpenAI's rival, Google, recently released its own Gemini 2.0 model earlier this month.

OpenAI, AI, models