OpenAI has introduced the o1 model, alongside the addition of the o1 pro mode, allowing the model to take longer to answer questions. This approach enables o1 to excel in challenging tests such as AIME 2024, which includes competitive math problems, programming tasks, and advanced scientific inquiries. The key feature is the model’s accuracy in providing correct answers every time, not just occasionally, showing a significant improvement.
An example from the programming test on Codeforces shows that the original o1-preview achieved a score of 62%, but after four reruns with all correct answers required, the score dropped to only 26%. On the other hand, o1 decreased from 89% to 64%, while o1 pro mode reduced from 90% to 75%, indicating that the pro mode consistently performs better and more reliably.
The ChatGPT Pro account, priced at $200 per month, allows unlimited usage of o1 and o1-mini (with separate usage policies) and limited usage of o1 pro mode.
TLDR: OpenAI introduces the o1 model and o1 pro mode, enhancing performance in challenging tests and providing accurate answers consistently. The pro mode outperforms the standard mode, and the ChatGPT Pro account offers unlimited access to o1 and o1-mini with limited access to o1 pro mode.
Leave a Comment