DeepSeek’s V4 Pro series remains the active flagship reasoning model on its API as of mid-October 2026, with the August 13 general-availability checkpoint (V4-Pro-0813) still served separately despite the September 10 V4.1 Flash launch that introduced a new MoE architecture and native vision. The planned temporary reroute of V4 Pro traffic to Flash was reversed, leaving traders focused on when the promised V4.1 Pro upgrade will ship. DeepSeek’s recent cadence—roughly every 60 days—points to a possible November window after the September Flash release, while unconfirmed reports of a larger-scale successor fuel speculation about agentic and coding gains. Key near-term signals include any changelog updates, benchmark leaks, or developer-conference timing that could clarify the next Pro checkpoint.