The revolutionary reasoning model from OpenAI, previously known as “Project Strawberry,” is now publicly available.
OpenAI’s clarification: “GPT Next” is not a new model.
The model, now dubbed “o1,” was launched on Thursday. Unlike its predecessors that relied on predictive text, o1 has the ability to “think” or process before generating a response. OpenAI, in its announcement, revealed that they have designed a new batch of AI models that take more time to contemplate before replying.
On the OpenAI website, a demo showcases a user asking o1 to form a grammatically accurate sentence without repeating letters. The model takes some time to process and finally comes up with a witty response, “go fix my bed,” after 39 seconds.
A different video presents Research Lead Jerry Tworek challenging o1 with a logic puzzle involving the ages of a prince and princess. The model systematically decodes the problem, comprehending the variables and figuring out the necessary equations to solve the problem.
TechWorld Speed
The demonstrations reveal that o1 excels at solving complex mathematical problems related to various fields like science, physics, and coding. OpenAI’s announcement further explains the model’s utility in the healthcare sector for annotating cell sequencing data, in physics for generating complex mathematical formulas for quantum optics, and for developers to build and execute intricate workflows.
However, the preview of o1 comes with certain restrictions. It lacks the ability to browse the internet or process uploaded files and images. For such tasks, OpenAI recommends using the GPT-4o model.
In the interest of safety, the U.S. and UK Safety Institutes were provided early access to o1 for testing, in line with the previous AI executive order from President Biden.
OpenAI’s o1 model is currently available to developers and ChatGPT Plus users for preview. It can be accessed through the ChatGPT app via the dropdown menu.
