Recent AI model releases have driven intense trader focus on the "Which math problems will AI solve in 2026?" market. OpenAI’s Astra variants and internal models have generated verified or Lean-formalized results on over 100 open problems since August, including substantial progress toward Navier-Stokes and multiple Erdős conjectures, while Meta’s Muse Spark produced papers resolving five open questions across probability, group theory, and optimization using standard chat interfaces. Anthropic’s Claude and Google DeepMind agents have raised bounds on the Riemann zeta function and solved additional Erdős problems. These advances reflect scaling inference compute, multi-agent systems, and formal verification tools rather than isolated benchmarks. With only weeks left before year-end resolution, traders weigh rapid lab momentum against strict criteria requiring official announcements of complete solutions and community verification for flagship problems like the Riemann Hypothesis or P versus NP.