Strawberries: OpenAI’s Next Step to Outperform Humans

OpenAI continues to work at full speed to improve the capabilities of its artificial intelligence, and in addition to GPT-5the next evolution of its AI is also reportedly implementing a code project called Strawberrydesigned to significantly improve AI’s reasoning abilities and enable autonomous exploration on the web.

In accordance with ReutersOpenAI’s Strawberry project aims to train AI to “plan ahead on the internet autonomously” and conduct “deep investigation.” The latter could help dispel hallucinations, as language models often resort to making up information when they lack training data on a particular topic.

Strawberries will become an evolution Project Q* from OpenAI, which had been previously rumored and touted as a significant step forward in artificial intelligence capabilities.

However, exactly how Strawberry works remains a closely guarded secret even within OpenAI, and there is no clear timeline for when it might become publicly available.

Recent report from Bloomberg Also confirming Strawberry’s existence, OpenAI is working on a secret project called Strawberry, which aims to achieve human-like thinking and improve AI’s deep research capabilities. While little is known about the internal details of Strawberry, it may have similarities to a method known as Self-Learning Intelligence (STaR). This method allows an AI model to improve its performance by learning from its own reasoning, creating and refining rationales for answering complex questions. This suggests that STaR could help AI models achieve greater-than-human intelligence in the long term.