OpenAI: ChatGPT maker announces o1 model with reasoning abilities

After rumors have been circulating for several months, OpenAI has finally released its ‘Strawberry’ series of AI models called o1.

This o1 model is what many in the technology world refer to as Project Strawberry. It’s the first of the series to be shared and was announced via a blog post on OpenAI’s website on Thursday (September 12).

The Microsoft-backed company says the series of AI models have been “designed to spend more time thinking before they respond. They can reason through complex tasks and solve harder problems than previous models in science, coding, and math.”

Before the announcement, the team conducted a series of tests with the team describing the “next model update” as being able to perform “similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology.”

The model was also found to excel in math and coding, with it being a major step up from GPT-4o. In a qualifying math exam, GPT-4o solved 13% of problems correctly while the reasoning model scored 83%.

The new model doesn’t yet have the features that make ChatGPT useful for everyday use, like browsing the web for information. The team says this is why they’ve reset “the counter back to 1″ and  are “naming this series OpenAI o1” as it represents a “new level of AI capability.”

An o1-mini version is being released for developers as OpenAI believes this will be a more efficient solution as it’s a faster and cheaper reasoning model that is effective at coding. This will be 80% cheaper than o1-preview.

OpenAI o1 solves a complex logic puzzle. pic.twitter.com/rpJbh8FkAg

— OpenAI (@OpenAI) September 12, 2024

Who will be able to use OpenAI’s o1?

To use OpenAI’s o1, either the preview or the mini version should be selected manually in the model picker.

Members who pay for ChatGPT Plus and Team gained access to o1 models on Thursday, with Enterprise and Edu users able to try it out from the beginning of next week.

Developers who qualify for API usage tier 5 can start prototyping with both models in the API with a rate limit of 20 RPM. The limits will be increased after additional testing.

There are plans to introduce the o1-mini to all ChatGPT users, including those on the Free plan.

Featured Image:  Via OpenAI blog

The post OpenAI: ChatGPT maker announces o1 model with reasoning abilities appeared first on ReadWrite.