OpenAI announced "reasoning" models — o3 and o3 mini

OpenAI announced the testing of new artificial intelligence models named o3 and o3 mini, capable of solving more complex tasks. Language models can be tested by security researchers. The release of o3 mini is scheduled for January 2025.

On real programming tasks, SWE Bench, the o3 model achieved an accuracy of 71.7%, which is 20 percentage points higher than the performance of o1.

In the EpochAI Frontier Math mathematical test, o3 achieved an accuracy of 25.2%, which is more than 12 times better than previous models.

The o3-mini model supports three levels of reasoning: low, medium, and high. The more time the model spends processing a task, the more accurate the result. In low latency mode, o3-mini demonstrates speed on par with GPT-4o, allowing for quick responses.

The new o3 mini models are planned to be released by the end of January 2025, and the full version of o3 will appear after that.

Source: x.com