Google's leading 39.5% implied probability reflects its strong recent Gemini iterations delivering competitive math scores on Text Arena leaderboards, where U.S. and Chinese models now cluster tightly in Elo ratings and domain-specific benchmarks like AIME. Alibaba's 22.5% position stems from Qwen-series advances in reasoning chains and cost-efficient scaling that have narrowed historical gaps with Western labs. Traders weigh ongoing model updates, open-weight releases from DeepSeek and others, and the fluid nature of arena voting that can shift on new capability demonstrations before October resolution. Key swing factors include next benchmark drops and any major lab announcements on specialized math training.