**Chinese AI labs have narrowed the frontier gap to roughly 3–12 points on composite indices like Artificial Analysis (top Chinese scores 44–46 vs. US leaders at 53–58 as of mid-October 2026), driven by rapid releases of large open-weight models.** Xiaomi’s MiMo-V2.6-Pro, Alibaba’s Qwen3.8 Max, Zhipu’s GLM-5.3, Moonshot’s Kimi K3, and DeepSeek’s V4.1 Flash/Pro series lead domestic efforts, with notable gains in agentic coding, math, and long-context benchmarks—sometimes matching or exceeding US models on self-reported or task-specific tests—while offering far lower pricing and MIT-style licensing. US closed models from Anthropic and OpenAI retain edges in overall reasoning and verified benchmarks. With frequent Chinese iterations, potential larger-scale releases (e.g., Alibaba’s teased multi-trillion-parameter ambitions), and only weeks left until year-end, trader consensus reflects sustained momentum but highlights the short window and historical US cadence as key variables.