OpenAI has released a new generative AI model, OpenAI o1, it is the company’s first model with reasoning capabilities like a human would. Internally known as Strawberry, o1 is a family of models, two of them will be available from today: o1-preview and o1-mini. Here are the details.
New o1 AI Models: Details
OpenAI said in a blog post that it trained this new AI model to “spend more time thinking through problems before they respond,” much like a person would think.
o1 models take more time to answer as they can refine their thinking process, try different strategies, and even recognise mistakes and correct them while looking for an answer. This means that ChatGPT can answer complex reasoning questions (like math problems, complex codes, etc) more accurately, without making factual errors.
The company said that o1 models can perform like PhD students on challenging benchmark tasks in physics, chemistry and biology. These also excel in math and coding. While the company’s previous model GPT-4o correctly solved only 13% of problems in the International Mathematics Olympiad, the o1 models scored 83%.
Currently, o1 models on ChatGPT don’t have internet access. Files or images can’t be uploaded to converse with it either.
ChatGPT Plus and Team users can access o1-preview and o1-mini starting today, and Enterprise and Edu customers will get access to these starting next week. OpenAI says it will bring o1-mini to all free users of ChatGPT Free users, but a timeline isn’t known. It is also worth noting that these o1 models have a capping on ChatGPT, there’s a weekly rate of 30 messages for o1-preview and 50 messages for o1-mini.
Microsoft has announced that the o1-preview and o1-mini models are now accessible in Azure AI Studio and GitHub models to a limited number of Azure customers.
