OpenAI has introduced ChatGPT Pro, a subscription plan priced at $200 per month, aimed at providing enhanced access to its AI models and tools. This plan offers unlimited use of the most advanced model, OpenAI o1, along with o1-mini, GPT-4o, and Advanced Voice. Notably, it includes o1 pro mode, which utilizes more computational power for improved responses to complex problems.
ChatGPT Pro is designed for professionals like researchers and engineers, enabling them to leverage advanced AI for productivity. The o1 pro mode is particularly effective, yielding more accurate and comprehensive responses in fields such as data science, programming, and legal analysis. Performance evaluations show that o1 pro mode outperforms both o1 and o1-preview on challenging benchmarks in math, science, and coding.
Reliability Metrics
The reliability of o1 pro mode is highlighted through a strict evaluation criterion, where a model must achieve correct answers in four out of four attempts to be deemed successful. Performance data shows significant improvements in reliability across various assessments, including:
Competition Math (AIME 2024):
- o1 pro mode: 80% (4/4 reliability)
Competition Code (Codeforces):
- o1 pro mode: 75% (4/4 reliability)
PhD-Level Science Questions (GPQA Diamond):
- o1 pro mode: 74% (4/4 reliability)
User Experience
Pro users can access o1 pro mode through the model picker, with a progress bar indicating response generation time. Notifications will alert users if they switch conversations while waiting for an answer.
To foster advancements in beneficial fields, OpenAI is awarding 10 ChatGPT Pro grants to medical researchers in the U.S., with plans for future expansions. Recipients include notable researchers specializing in rare diseases, biomedical data, aging, and cancer immunotherapy.
OpenAI plans to enhance the Pro plan with additional capabilities for more compute-intensive tasks, while also making some features available to other subscribers.
