Traders see a tight contest for the top Text Arena math model by year-end, with Anthropic holding a modest edge at 36% implied probability over Google's 30%. Recent releases of reasoning-focused large language models from both labs have narrowed performance gaps on complex math benchmarks, driven by advances in chain-of-thought training and synthetic data techniques. OpenAI, xAI, and several Chinese developers trail at around 13%, reflecting competitive but less differentiated math capabilities to date. Key swing factors include upcoming model updates that could demonstrate superior handling of multi-step problems or novel architectures, alongside any third-party arena evaluations that reward demonstrated accuracy over marketing claims. High uncertainty stems from rapid iteration cycles typical in the sector.