News on October 28: recently, attentive netizens discovered that DeepSeek has once again switched to a brand-new large model fingerprint, triggering widespread speculation that it is about to release a major new model. At present, brand-new code including DeepSeek-V5, V5-open and V5-next has formally appeared in the official DeepSeek Harness code repository.
Judging from the code structure, the current V5-related content mainly serves as a placeholder, but this clearly shows that the release mechanism for the new model is ready. When the official launch comes, the team will only need to modify the /models configuration on the server side to push out V5 directly, without forcing clients to release a new version. Specifically, the code system covers two entirely different delivery paths: deepseek-v5-open takes the version-threshold-free route, while deepseek-v5-next belongs to the path that requires a client upgrade.
Of course, the existing code only proves that DeepSeek has done the groundwork for releasing the V5 series of large models, and does not mean the series will be launched immediately. Looking back at its recent cadence, V4.1Flash, released on September 17, adopted a completely new architecture but still carried the 4.0 series in its naming. Considering the feedback when the previous-generation Pro official version was released, it is also a reasonable extrapolation that the officials will directly name the new-generation flagship large model DeepSeek V5Pro in the future.
Industry analysis holds that, given that the Flash version was only released recently, the V5Pro series will most likely not go live immediately within the 10th month (October); it is expected to officially meet the public at the earliest in the 11th month (November). If progress goes smoothly, in mid-11th month (mid-November), V5Pro and V5Flash may hopefully receive updates together, making the product naming sequence clearer.