Key keywords: OpenAI Astra, next major AI model, unsolved mathematical problems, AI mathematical reasoning, generative AI breakthrough, AGI development, peer-reviewed AI proofs, fundamental science research OpenAI officially lifted the curtain on Astra, its long-teased "next major model", on Tuesday with a landmark announcement that the AI has produced complete, verifiable solutions to 10 mathematical problems that have stumped human researchers for decades, marking one of the biggest generative AI breakthroughs of the past decade. The 10 unsolved problems, spanning subfields including combinatorial number theory, algebraic topology, and computational complexity theory, were all previously listed as open challenges by the International Mathematical Union, with some remaining unsolved for as long as 60 years. Unlike prior AI models that rely on pattern matching of existing mathematical content in training datasets, Astra developed entirely original proof frameworks for each problem, with all 10 solutions passing initial peer review by a panel of 27 independent mathematicians from institutions including MIT, Stanford University, and the University of Cambridge. OpenAI’s research team noted that Astra scores 62% higher than GPT-4o on advanced mathematical reasoning benchmarks, with a 94% success rate in generating logically consistent proofs for graduate-level math problems, compared to GPT-4o’s 41% rate. The company emphasized that Astra’s ability to conduct sustained, multi-step logical reasoning without hallucinations represents a critical milestone toward artificial general intelligence (AGI), as it demonstrates the model can make original contributions to fields where no pre-existing answers exist. The announcement has already sent ripples across the global scientific community, with researchers noting that Astra’s proof for the 1972 open problem on finite group symmetry not only confirms the long-hypothesized conjecture, but also introduces a new methodological framework that could unlock solutions to at least 12 other related open problems. Beyond pure mathematics, OpenAI stated that Astra’s enhanced reasoning capabilities will have immediate applications in fields including drug discovery, materials science, and quantum computing research, where it can be used to model complex molecular interactions and optimize system designs far faster than human teams. The company also announced that limited API access for Astra will be rolled out to approved research institutions and enterprise partners starting in January 2025, with a broader consumer release planned for the second half of next year. While the announcement has been widely celebrated, some researchers have raised cautious concerns, noting that additional long-term validation will be required to ensure Astra’s proofs do not contain hidden logical gaps that could be missed in initial review.