On October 28, attentive users recently discovered that DeepSeek has once again adopted an entirely new large-model fingerprint, prompting widespread speculation about the imminent launch of a significant new model. At present, the official DeepSeek Harness codebase has officially added new code, including DeepSeek-V5, V5-open, and V5-next.
From the perspective of the code structure, the current V5-related content primarily serves as placeholders, yet this clearly demonstrates that the mechanism for launching the new model is already prepared. When the time comes for an official release, the team will only need to modify the /models configuration on the server side to directly roll out V5, without requiring clients to update their versions. Specifically, the code system covers two distinctly different distribution paths: deepseek-v5-open follows a version-free approach, while deepseek-v5-next requires clients to upgrade.
Of course, the existing code merely confirms that DeepSeek has laid the groundwork for the release of the V5 series of large models; it does not mean that this series will be launched immediately. Looking back at its recent pace, although the V4.1Flash released on September 17 featured a completely revamped architecture, its naming convention still retained the 4.0 series. Considering the feedback received upon the official release of the previous Pro version, it is reasonable to speculate that the company may eventually name the next-generation flagship large model DeepSeek V5Pro.
Industry analysts believe that, given how recently the Flash version was released, the V5Pro series is unlikely to go live within October. It is expected that the official public unveiling will not occur until at least November. If all goes smoothly, by mid-November, both V5Pro and V5Flash could potentially receive updates simultaneously, resulting in a clearer product naming sequence.