OpenAI may release its most powerful AI model in September-November 2024 and could integrate it into ChatGPT-5. OpenAI was working without publicity on project Q* or Q-star. The same has now been code-named Project Strawberry.

This project is capable of doing autonomous internet research. It will improve its AI reasoning capabilities dramatically. It is a step in the direction of AGI.

Sam Altman, OpenAI CEO, posted an image of strawberries growing in two pots on his X account (August 7, 2024). It confirms that the work on the new project is in progress.

OpenAI has already demonstrated the model to the national security officers as a part of its commitment to transparency.

The Information, the tech industry publication, reported on August 26, 2024, that project strawberry would be better at math and programming. Once integrated into ChatGPT, it would become the most powerful AI chatbot. According to the Information, the new model is capable of advanced level of thinking. It would be able to solve puzzles. It may try its hand on Connections, the NYT word puzzle.

The next model in the pipeline is Orion, and project Strawberry would facilitate the raising of capital for this next generation model. As the training data has been exhausted, the future models such as Orion could use synthetic data. OpenAI is testing how to generate high-quality synthetic data in large amounts. Such data can fill gaps in real world data sets. These could make the models more neutral and fairer. The accuracy of the models would go up.

Project strawberry could conduct experiments autonomously, analyze data and come up with new hypotheses. This could lead to new breakthroughs (including the new drug discovery). The models can customize education, create educational content and interactive lessons.