OpenAI Launches "Strawberry" AI Models to Solve Complex Problems with Reasoning Capabilities.

OpenAI has unveiled its latest series of artificial intelligence models, codenamed "Strawberry," designed to handle more complex tasks with greater precision. These models, announced on Thursday, represent a significant leap in AI reasoning capabilities, particularly in areas such as science, coding, and mathematics.

Dubbed "o1" and "o1-mini," the new models are available starting Thursday through ChatGPT and OpenAI's API. According to a blog post from the AI company, these models spend more time processing responses to deliver more accurate solutions to difficult queries. The announcement follows a Reuters report which first revealed details about the models.

OpenAI Launches "Strawberry" AI Models to Solve Complex Problems with Reasoning Capabilities.

Noam Brown, a lead researcher at OpenAI specializing in improving the reasoning abilities of its AI models, confirmed on the social media platform X (formerly Twitter) that the models originated from a project known internally as "Strawberry."

"I'm excited to share with you all the fruit of our effort at OpenAI to create AI models capable of truly general reasoning," Brown wrote.

The o1 model is proving its worth in challenging fields. It scored 83% on the International Mathematics Olympiad qualifying exam, a remarkable jump from the 13% achieved by OpenAI's earlier model, GPT-4o. The model also demonstrated significant improvement in tackling competitive programming challenges and surpassed human PhD-level accuracy in a benchmark of science-related problems.

A standout feature of the Strawberry models is their use of "chain-of-thought" reasoning, a method that breaks down complex problems into smaller, logical steps. This approach, which has been shown to improve AI performance, is now fully automated in the o1 and o1-mini models, allowing them to independently solve intricate problems without external prompts.

"We trained these models to spend more time thinking through problems before they respond, much like a person would," OpenAI explained. "Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes."

OpenAI’s Strawberry project was first reported by Reuters in November 2023 under the code name Q*, with its official title revealed earlier this year.