Google and Anthropic currently dominate the Text Arena Math leaderboard with Gemini 4 Argon high and Claude Fable/Opus 5 variants posting the top lab scores near 1525–1530, driven by recent proprietary model releases emphasizing multi-step reasoning on math prompts. OpenAI’s early-October release of hundreds of AI-generated proofs on longstanding problems has lifted its market-implied odds but has not yet translated into comparable arena performance. Meta and several Chinese labs trail closely enough that incremental updates before year-end could reorder second place, while the narrow spreads among leaders underscore how small benchmark gains in large language model capabilities can shift trader consensus on end-of-December positioning.