DeepWhale Mobile Releases Android APK: DeepSeek Harness on Your Smartphone

⚡ The gist in 5 seconds - An Android APK of the “ Mobile” app — a mobile harness for DeepSeek — has been published in the feely0208/deepwhale-mobile repository. - The build can be grabbed directly from GitHub; the HarmonyOS version is still going through (app registration). - Supported models, device requirements, and the method of connecting to DeepSeek are not disclosed in the project description. ### 🔍 What was found A repository called feely0208/deepwhale-mobile has appeared on GitHub with a project named “ Mobile” (Deep Whale Mobile). Its description positions it as a “DeepSeek Harness on the phone”: the Android APK is already published, while the HarmonyOS build is at the stage — the mandatory registration of an app with the Chinese regulator before distribution via app stores like AppGallery. The project author is user feely0208, meaning this is a third-party development, not an official DeepSeek client. The word “harness” in the description points to a wrapper around the model: a layer through which the app sends requests to DeepSeek and manages them. How exactly this is implemented in “ Mobile” — whether it calls a cloud API, whether local model weights are supported, and which model versions are claimed — does not follow from the repository: the description is limited to the name, build status, and a link to the project. There is also no data on the license, APK size, or publication date. ### 💡 Why it matters Mobile access to DeepSeek today is covered either by the official app or by third-party clients, and the appearance of yet another Android harness expands the options for those who work with DeepSeek models from a phone. A separate signal is the preparation of a HarmonyOS build: this shows the author is targeting not only global Android but also the Chinese market, where is a mandatory condition for distribution. A practical caveat: the APK is installed manually, outside Google Play, and the source code and claimed capabilities are not documented in any way — before installing, it is worth reviewing the repository code, since an unverified build gains access to your conversations with the model.