DeepSeek's August 13, 2026 general availability release of V4 Pro 0813, which upgraded agentic benchmarks like DeepSWE and Terminal-Bench through targeted post-training on the 1.6T MoE architecture, established the current flagship. The September 10 launch of the smaller V4.1 Flash model with native vision and a new causal encoder-decoder design temporarily routed Pro API traffic, but the company reversed that plan and kept V4 Pro live without announcing a V4.1 Pro successor. As of mid-October 2026, no official timeline, changelog entry, or Hugging Face weights exist for the next Pro variant despite DeepSeek's monthly cadence on lighter models and rapid Chinese AI competition. Traders appear to price in extended uncertainty around flagship iteration timelines and potential surprise announcements.