OpenAI has unveiled its new major product: an AI model code-named Strawberry, officially known as OpenAI o1. More specifically, o1 represents a family of models rather than just a single model. This new series of AI models includes: o1-preview and o1-mini. The o1 models have been designed to spend more time thinking through problems before delivering responses, making them especially effective for complex tasks in fields like science, coding, and mathematics.
“We trained these models to spend more time thinking through problems before they respond, much like a person would. Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes,” OpenAI said.
OpenAI claims that during testing, the o1 models performed at the level of PhD students on challenging benchmark tasks in physics, chemistry, and biology. For instance, in a qualifying exam for the International Mathematics Olympiad, o1 outperformed GPT-4o, solving 83% of the problems compared to GPT-4o’s 13%.
Also read: OpenAI unveils cost-efficient GPT-4o mini model: What is it and how it works
One key feature of the o1 models is their advanced safety measures. OpenAI has come up with a new safety training approach that harnesses their reasoning capabilities to make them adhere to safety and alignment guidelines. In tests to prevent “jailbreaking” (where users try to bypass rules), o1-preview scored much higher than previous models.
Now, let’s talk about the o1-mini. It is a smaller and more affordable version of the o1 model. It focuses on coding tasks, making it a cost-effective option for developers. While o1-preview is more powerful, o1-mini offers faster response times and is 80% cheaper, making it ideal for projects that require advanced reasoning without needing extensive world knowledge.
Also read: OpenAI planning to increase subscription prices as it continues to bleed money
Starting today, ChatGPT Plus and Team users can access both o1-preview and o1-mini models, with plans to expand access to ChatGPT Enterprise and Edu users next week. OpenAI is also planning to bring the o1-mini access to all free users. Developers, who qualify for API usage tier 5, can also start working with these models through the OpenAI API.
Also watch: