OpenAI’s September 8 announcement that an internal model coordinated 10,000 AI agents to produce a Lean-verified solution to the Navier–Stokes existence and smoothness Millennium Problem has sharply elevated expectations for near-term AI progress on FrontierMath open problems. The company followed with claims that a newer internal model solved over 100 longstanding math problems, including a second Millennium Problem, after roughly one month of training. OpenAI’s GPT-5.6 and GPT-6 Astra variants currently lead FrontierMath v2 leaderboards at 84–89 percent on private tiers, outpacing Anthropic’s Claude models. Epoch AI’s official classification of any additional Open Problems entry as solved remains the sole resolution trigger for related markets, so traders are watching for formal verification of the Navier–Stokes proof and any new Epoch AI updates through late September.