OpenAI Advances New Reasoning Technology Dubbed ‘Strawberry’
OpenAI is actively developing an innovative approach to enhance its artificial intelligence models under the project codename “Strawberry,” according to sources familiar with the matter and internal documents reviewed by Reuters. This previously undisclosed initiative highlights OpenAI’s commitment to advancing advanced reasoning capabilities within its models. Teams within OpenAI are currently immersed in the Strawberry project, as outlined in recent internal documents from May. While details remain guarded, ongoing research efforts indicate significant strides towards overcoming challenges in AI, such as planning and multi-step problem-solving.
Strawberry is designed not only to answer queries but also to autonomously navigate the internet for comprehensive research, showcasing OpenAI’s ambitious goals. The project aims to bolster AI models’ reasoning capabilities through specialized post-training methods, akin to Stanford’s Self-Taught Reasoner (STaR), enabling iterative intelligence enhancement.
OpenAI intends to unveil Strawberry with enhanced reasoning capabilities, engaging developers and stakeholders outside the company. The technology promises to tackle long-horizon tasks and perform complex actions autonomously over extended periods, leveraging deep-research datasets. Ultimately, OpenAI seeks to surpass current AI limitations, potentially transforming fields from scientific research to software engineering.