.jpg)
In September 2024, OpenAI introduced the o1 model, which brought reasoning support, allowing it to answer more complex questions by thinking more deeply. Now, about three months later, OpenAI today announced the introduction of their new model, namely o3.
However, the o3 model has not been offered to the public yet. It is open to researchers first, especially those who study the security of artificial intelligence models. According to the initial target, the o3-mini model is expected to be offered around the end of January 2025.
Like the o1 model, the use of the o3 model also takes some time in providing feedback to questions. This allows it to make checks, and ensure that it offers the best answer to users. However, like the normal hallucinatory model, several researchers have previously found that models with reasoning support can also deceive humans through the answers given.
Some early testers also found that the o3 model also had some problems in responding and thinking like a human - saying it is still a long way from reaching the AGI phase.