DeepSeek’s September 10, 2026 launch of V4.1 Flash—the first model in a new Causal Encoder-Decoder MoE architecture with native vision, 1M-token context, and strong agentic coding and tool-use benchmarks—has shifted focus to the expected V4.1 Pro follow-up. The lab maintains a rapid cadence of roughly one major release every 20–60 days, with the prior V4-Pro-0813 reaching general availability on August 13 after an April preview. V4.1 Flash currently routes some Pro-tier traffic while the company expands partnerships (including Huawei Ascend tooling) and prepares for a potential IPO. No official date or confirmation exists for V4.1 Pro, though trackers project a November window; traders watch for changelog updates, benchmark leaks, or executive statements that could accelerate or delay the flagship reasoning model.