OpenAI is reportedly developing a new AI project codenamed ‘Strawberry’ that aims to solve complex mathematical problems and optimize programming tasks. This project, previously known as ‘Q*’, is expected to be unveiled this fall and could potentially enhance ChatGPT by incorporating its advanced capabilities. Sources claim that ‘Strawberry’ is capable of achieving scores exceeding 90% on benchmark tests for complex mathematical problems, significantly surpassing GPT-4’s performance. This advancement is fueled by the ability of ‘Strawberry’ to effectively ‘think’ over time and solve problems with greater efficiency. OpenAI is also working on a flagship language model named ‘Orion,’ designed to surpass even GPT-4, with ‘Strawberry’ providing valuable training data for this ambitious project. This combination of ‘Strawberry’ and high-quality synthetic data is expected to boost ‘Orion’s’ accuracy and minimize errors, marking a significant leap forward in AI technology.