DeepSeek V5 series appears in official code repository, placeholders already in place before release
Placeholder code such as DeepSeek-V5, V5-open, and V5-next has appeared in the official DeepSeek Harness code repository, meaning the release mechanism for the V5 series is ready; in the future, the service side can go live by simply changing /models, with no need for a client release. Among these, V5-open follows a path with no version threshold, while V5-next requires a client upgrade, covering both distribution paths. The code does not mean V5 will be released soon, but groundwork for the release has begun.
Topics
Body · AI translation
In the official code repository of DeepSeek Harness, some new code has already appeared, including DeepSeek-V5, V5-open, V5-next, and so on. It looks like this is the next generation of large models.
What exactly does this code mean? You can't just see V5 and make a big news story out of it. Ordinary people who don't understand code just ask Da Fei Yu directly, and the explanation given is that the mechanism for the new model is ready, but the current v5 is a placeholder.
The significance of this design is: DeepSeek can change /models on the server side to roll out v5, and the client does not need to release a new version. deepseek-v5-open follows the "no version threshold" path, while DeepSeek-v5-next follows the "client upgrade required" path, which exactly covers both distribution paths.
To put it plainly, DeepSeek official has already made some advance preparations for the release of the DeepSeek V5 series of large models, and is now using placeholders to occupy the position. After the future release, the server only needs to change /models to release it, and the client will not need to release a new version afterward to support it.
The current code does not show that DeepSeek will release the V5 large model very soon, but it does prove that it is getting closer and closer, and that groundwork has begun for the release.
DeepSeek's naming of large models has also always been rather confusing. The V4.1 Flash released on September 17 has already completely switched to a new architecture and is no longer comparable to the previous V4, yet the naming still continued the 4.0 series. The version released on 1017 also brought great improvements, but it still did not become V4.2.
Considering the not-so-pleasant memories of the previous V4 Pro official version, it would also be a good approach for DeepSeek to directly change the new-generation flagship large model to DeepSeek V5 Pro. The Flash version has not been released for long, but V5 Pro will not be released very soon either. This month is probably out of the question; it will have to be November.
If it is mid-November, then DeepSeek V5 Pro and V5 Flash could just switch to the new generation together, and the naming would not be so confusing.