Google AI Edge Gallery, a new app for Android that allows users to experience running artificial intelligence (AI) models locally on a device, was released on Sunday. The Mountain View-based tech giant says the experimental app can find, download, and run large language models (LLMs) entirely on-device without requiring an Internet connection. Currently, only a limited number of use cases are available via the app, including chatting, image analysis, as well as a few text-based and coding tasks. The company said that an iOS version of the app will soon be released.
Google AI Edge Gallery App Is Available on GitHub
The app is currently available as an “experimental alpha release” on the tech giant’s GitHub listing. The company has provided an Android application package (APK) file that can be downloaded, as well as a detailed guide to installing it on an Android device. The app, which is 115MB in size, is available with a permissive Apache 2.0 license, which allows both academic and commercial use cases.
Gadgets 360 staff members were able to download and install the app without any hassle. If you are new to installing APK files, Google has provided a detailed guide here. However, do note that downloading and installing APK files is always a risky move since these apps are not verified by a trusted app marketplace and might contain malware. Notably, we did not find any malware or viruses in this app.
The Google AI Edge Gallery app comes with a list of AI models that users can download and run locally on their devices. This list of models will vary depending on how new the device is and whether it has an AI-enabled chipset or not. Additionally, the app also lets users import and run a model that is already downloaded on the device.
There are three primary features users can explore with this app. First is Ask Image, which is an image analysis feature. After downloading and running a model, users will be able to upload an image and ask the AI questions about it. Second is AI Chat, which lets users have a conversation with the model. Since the app only supports local models, these models may not have up-to-date knowledge.
Finally, the third feature is Prompt Lab, which is a space with several AI-powered features such as tone-based rewriting, text summarising, free-form generation, as well as code snippet generation. Notably, while running any AI model, users will be able to configure aspects such as tokens, temperature, accelerator, and more. Users can also check the benchmark metrics of the model.
