**Alibaba and Xiaomi lead the closely matched market-implied odds at 36% and 34.3%, respectively, reflecting Xiaomi’s recent MiMo-V2.6-Pro topping several September 2026 Chinese model benchmarks alongside strong OpenRouter token share, while Alibaba’s Qwen3.8-Max benefits from unmatched Hugging Face downloads, broad model family, and enterprise/cloud distribution.** Moonshot’s earlier Kimi K3 release positions it at 12% for long-context agentic coding strengths, with Z.ai (GLM) close behind at 11.2% on coding benchmarks, but both pure-plays face recent regulatory scrutiny over data practices that has weighed on sentiment. DeepSeek’s efficiency edge and high valuations have not translated into higher odds amid the same probes. Key swing factors before end-October include any new model drops, benchmark updates, or resolution of the data-security investigations, as Chinese labs continue rapid iteration across open weights, cost, and multimodal capabilities in a fragmented field without a single dominant leader.