OpenAI’s early October 2026 release of 722 papers from an internal frontier model—claiming solutions or substantial progress on 377 previously open problems across number theory, algebra, topology, and Millennium Prize areas like the Navier-Stokes equations and quasi-Riemann hypothesis—has sharply lifted trader expectations. Many proofs include Lean formalizations, with average compute equivalent to three hours of ChatGPT Pro usage, prompting rapid Epoch AI review of FrontierMath Open Problems. Competing labs such as Anthropic (via Claude models on FrontierMath v2 leaderboards) and Google DeepMind (AlphaProof) continue advancing formal verification benchmarks, while GPT-6 Astra variants lead private tiers at over 93 percent. Key near-term catalysts include Epoch AI’s next official classifications of solved problems and any follow-on releases or Lean-verified submissions before year-end deadlines.