OpenAI launches ‘strawberry’ AI models to tackle complex prompts
OpenAI, backed by Microsoft, has unveiled its latest "Strawberry" series of AI models, which are intended to spend more time processing answers in order to handle more complex and challenging queries.
According to OpenAI, the new models, known as o1 and o1-mini, can reason through complex tasks and outperform previous models in fields such as science, coding, and mathematics.
The project, internally codenamed "Strawberry," was first reported by Reuters. OpenAI announced that the o1 model would be available in ChatGPT and via its API beginning Thursday. Noam Brown, an OpenAI researcher who specialises in model reasoning, confirmed the Strawberry project's release on the social media platform X, expressing his excitement about the models' improved reasoning capabilities.
In a blog post, OpenAI highlighted the o1 model's impressive performance, which scored 83% on the International Mathematics Olympiad qualifying exam, compared to only 13% for its predecessor, GPT-4o. The model also outperformed human PhD-level accuracy on scientific benchmarks and significantly improved in competitive programming tasks.
These models' improved reasoning abilities are due to a technique known as "chain-of-thought" reasoning, which divides complex problems into smaller, more logical steps. OpenAI has automated this method, allowing models to approach and solve complex problems without user intervention.
"We trained these models to spend more time thinking through problems before responding, similar to how a person would," the company said. "This process helps them refine their strategies, recognise mistakes, and ultimately improve their problem-solving abilities."
Leave A Comment