OpenAI's Astra: A Breakthrough in Solving Mathematical Problems
This model is set to be released soon.
The Astra neural network, developed by OpenAI, has successfully found solutions to 10 mathematical problems that have long eluded human researchers. Some of these issues have been known for decades.
Astra is a multi-agent system and the next model from OpenAI, currently preparing for release. Notably, the model was able to not only provide proofs for the problems but also formalize its solutions in Lean.
Read this article in full and everything else in Telegram, ad-free
Open in TelegramThanks, I just want to finish reading here
The solution was relatively inexpensive for OpenAI. The company does not disclose specific figures, but it is known that the number of tokens used by Astra for solving would have cost around $2000 for the current top model GPT-5.6 Sol.
Together with the Lean proofs, the company published a description of the problem-solving process and a scientific paper dedicated to the solutions. The work was not written by the neural network itself; it was prepared by the company's staff.
The release date for the Astra model has not been announced. It is expected that this neural network will be the first OpenAI development released after thorough testing by U.S. authorities.
