An internal iteration of OpenAI’s Astra model has successfully solved 10 major open math and computer science problems. This advancement signals a significant leap in the reasoning capabilities of large language models.