Recent successes by OpenAI’s GPT-6 Astra, which solved multiple high-tier FrontierMath problems including a “major progress” level conjecture in September 2026, have driven strong trader consensus toward near-term resolutions on the Open Problems benchmark. Anthropic’s Claude Fable 5 and Fable 5.1 models also contributed verified solutions earlier in the year, such as the 2-adic absolute Galois group problem, while Epoch AI’s updated policy on human-AI collaboration and FrontierMath Erdős benchmark (where Astra scored 3% under fixed budgets) highlight accelerating capabilities. Competitive dynamics favor OpenAI’s lead in research-level math, with Tier 4 nearing saturation at reported 97.6%. Traders watch for Epoch AI’s next official classifications on remaining open problems, expected before year-end deadlines, amid rapid model releases and verifiable Lean proofs.