According to Kuai Technology on September 13, OpenAI officially launched the OpenAI o1 preview model early this morning.

For complex reasoning tasks, the new model represents a new level of AI capabilities, characterized by spending more time thinking before answering, similar to the process of humans thinking and solving problems.

OpenAI has explained that GPT-4, released in 2023, is similar to the intelligence level of a high school student, while GPT-5 represents the growth of AI from a "high school student to a Ph.D. student," and the o1 model is a crucial step in this process.

According to tests, the o1 model achieved a score of 83% in the qualification examination of the International Mathematical Olympiad, while GPT-4o could only correctly solve 13% of the problems.

In the programming ability competition Codeforces, the o1 model achieved a percentile score of 89%, while GPT-4o only achieved 11%.

OpenAI stated that based on tests, in the next updated version, AI's performance in challenging benchmarks in physics, chemistry, and biology would be comparable to that of Ph.D. students.

Compared to existing large models like GPT-4o, OpenAI o1 can solve more difficult reasoning problems while improving the mechanical defects present in previous models.

For example, when answering programming questions, it will be more organized, thinking through the entire answer process before starting to write the code.

o1 spends more time thinking

In the o1 model, AI continuously trials and refines answers, significantly improving the accuracy and quality of generated results.

OpenAI announced that ChatGPT subscribers can access these two new models starting September 12, with a weekly message limit of 30 for o1-preview and 50 for o1-mini.