Alibaba's Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts model with 95 billion active parameters and 1M-token context, launched via API on August 3, 2026, with open weights following shortly after and a September 0902 API snapshot update. This remains the current flagship, topping internal benchmarks in coding, agentic tasks, and multimodal reasoning while competing with models like Moonshot's Kimi K3. Trader sentiment reflects Alibaba's rapid release cadence of smaller variants but slower flagship Max iterations, reinforced by the September announcement that Qwen 4 is in training with future versions targeting 5–10 trillion parameters. No verified preview or release timeline for a 3.9+ Max has surfaced in the past 30 days, pushing market-implied odds for an October delivery well below 50 percent amid ongoing competitive pressure from other Chinese labs.