Recent benchmark leadership by Xiaomi’s MiMo-V2.6-Pro, Alibaba’s Qwen3.8 Max, and Moonshot’s Kimi K3 has created a tight race for second-best Chinese AI company, with these labs posting scores within a few points on Artificial Analysis and BenchLM indices as of mid-October 2026. DeepSeek’s September V4.1 Flash release drove high OpenRouter volume and narrowed the U.S. gap to roughly 3 percent, yet its edge in cost and reasoning is matched by peers’ open-weight momentum and enterprise traction. Z.ai’s GLM-5.3 and StepFun’s upcoming Step 5 Preview add further volatility ahead of year-end resolution. Trader sentiment reflects these shifting capability and adoption signals rather than any single frontrunner.