OpenAI presented a new series of AI models for solving complex problems.

OpenAI presented a new series of artificial intelligence models o1

OpenAI has announced the release of a new series of artificial intelligence models called o1. The first model in this series, o1-preview, is already available in ChatGPT and the company's API. These new models have learned to spend more time thinking about tasks before answering, just as a human does. They are able to ponder complex tasks and solve more challenging problems in the scientific, software, and mathematical fields compared to previous models.

Test Results

According to the test results, the next model o1 demonstrated results at the graduate level in difficult tasks in physics, chemistry, and biology. It also showed high results in mathematics and programming. In the qualifying exam of the International Mathematical Olympiad, the o1 model correctly solved 83% of the tasks, while GPT-4 solved only 13% of the tasks.

o1-mini - a faster and cheaper version of the model

Additionally, a version of the model o1-mini, which is faster and cheaper, was presented. This version is optimized for programming and costs 80% less than the original o1-preview model.

Access to new models

At the moment, access to the new models has been granted to ChatGPT Plus and Team users. Developers with the appropriate level of API access can also start testing them. The company intends to expand access to these models soon.

Mandatory updates and model development

OpenAI notes that these models are early versions and regular updates and improvements are planned. The company also continues to develop the GPT series models along with the new o1 series.