Alibaba’s rapid 2026 cadence, highlighted by the August 26 release of the multimodal Qwen3.8-Flash-Next (a 125B MoE model activating just 6B parameters per token with Gated DeltaNet and Qwen Sparse Attention), drives trader sentiment around a near-term 3.9+ Flash variant. This open-weight checkpoint serves as an explicit architecture preview for Qwen4, delivering strong coding and agentic benchmarks at roughly one-ninth the training cost of prior Plus models while supporting up to 1M-token context. Competitive pressure from DeepSeek and Kimi, combined with Alibaba’s push for cloud monetization and developer ecosystem growth, reinforces expectations of continued iteration. Qwen4 remains in training with no confirmed 3.9+ Flash timeline, though historical patterns suggest potential movement before year-end.