Anthropic holds the market edge at 54.5% implied probability due to its sustained strength on the Text Arena Math leaderboard, where Claude variants like claude-fable-5-high post competitive Elo scores near 1524 amid human preference battles. Google trails at 32% as its Gemini-4-argon models currently lead some snapshots with scores around 1528-1530, reflecting rapid iteration in large language model reasoning benchmarks. OpenAI sits at just 6.5% despite October's release of hundreds of internal-model math proofs, as those gains have not yet shifted live arena rankings. With resolution weeks away, upcoming model drops from leading labs and continued voting could tighten or flip the narrow gap between the top two contenders.