Sept 13 V7N - Microsoft-backed OpenAI has unveiled its new "Strawberry" series of AI models, designed to spend more time processing queries to tackle complex problems. These models, internally code-named "Strawberry" and now called o1 and o1-mini, are built to excel in challenging tasks across science, coding, and mathematics. Starting Thursday, the o1 model will be available through ChatGPT and its API.

OpenAI researcher Noam Brown highlighted that these models are capable of general reasoning, a key goal of the Strawberry project. The o1 model has shown significant improvements, including scoring 83% on the International Mathematics Olympiad qualifying exam, compared to 13% for GPT-4o, its predecessor. The model also demonstrated enhanced accuracy in science benchmarks and competitive programming tasks.

The key to this leap in performance is a method called "chain-of-thought" reasoning, where the AI breaks down complex tasks into smaller, logical steps. This approach, traditionally used as a prompting technique, is now automated, allowing the models to independently refine their problem-solving strategies.

According to OpenAI, these advancements mean the models are trained to think through problems more thoroughly, much like a human, improving their ability to reason, adapt, and learn from mistakes. OpenAI's work on this project, initially reported as Q* in November 2023, was later renamed Strawberry.

END/TEC/RH