“A new generation of AI models that takes more time to think before responding. These models are capable of reasoning through complex tasks and tackling more challenging problems in science, coding, and math compared to their predecessors.”
New model update demonstrates performance on challenging benchmark tasks in physics, chemistry, and biology comparable to that of PhD students. It shows outstanding abilities in math and coding, with an impressive 83% score on the International Mathematics Olympiad (IMO) qualifying exam, compared to only 13% by GPT-4o.
Additionally, the model achieved the 89th percentile in Codeforces coding contests. While this early version lacks some features found in ChatGPT, such as web browsing and file uploads, it marks a significant leap in AI capability, especially for complex reasoning tasks. Consequently, we are initiating a new series with this model, named OpenAI o1.