OpenAI, led by Sam Altman, has unveiled the highly-awaited o1-preview AI models which signify a major advancement in the field of artificial intelligence, particularly for ChatGPT. These cutting-edge models excel in tackling intricate problems across various domains like science, programming, and mathematics.
The introduction of the o1-preview models marks a significant breakthrough, with a focus on enhanced reasoning capabilities resulting in improved problem-solving skills. Notably, in initial assessments, the updated reasoning model demonstrated performance comparable to that of PhD students in tasks related to physics, chemistry, and biology, while also displaying remarkable proficiency in math and coding competitions.
For instance, the o1 model scored an impressive 83% in a qualifying exam for the International Mathematics Olympiad, outshining the GPT-4o model by a significant margin. Despite lacking certain practical features like web browsing and file uploading present in GPT-4, the o1-preview model is positioned as a valuable asset for industries requiring sophisticated problem-solving abilities.
OpenAI has also unveiled a more compact and budget-friendly version named o1-mini, tailored to cater specifically to developers seeking advanced coding functionalities without the extensive knowledge base. This cost-effective solution is priced 80% lower than the o1-preview model, making it an attractive option for developers.
Moving forward, users of ChatGPT Plus and Team can now manually select either the o1-preview or o1-mini models for usage. While the o1-preview model has a rate limit of 30 messages, the o1-mini version allows for up to 50 messages. Additionally, API users in the highest usage tier are granted access to these models.
Despite certain limitations such as the absence of features like function calling and streaming, OpenAI is focused on ensuring the safety and security of these models. The company has introduced new security training methodologies, with the o1-preview model showcasing superior performance in jailbreak tests compared to GPT-4o.
Looking ahead, OpenAI plans to expand the availability of o1-mini to ChatGPT Free users while continuing to enhance the features of the o1 series with additions like web browsing and file uploads. Altman acknowledges the imperfections of the models but remains optimistic about their potential, emphasizing the dawn of a new era in AI capabilities. Meanwhile, OpenAI’s latest funding signals their ambitious goal of achieving a $150 billion valuation.