Google Lance Ai Edge Gallery, a local AI that works without connection
Google tends to want to reassure users about shared data when using an online conversational artificial intelligence. Indeed, exchanging with Chatgpt or Google Gemini is not without risks vis-à-vis respect for privacy. To make up for this problem, Google has just put Ai Edge Gallery online, an application that allows the use of AI models locally … on the smartphone.
Use artificial intelligence in offline mode
Available since The GitHub platformthe new Google application is still at the experimental stage. His ambition? Allow you to run large languages of languages directly on your smartphone, in offline mode.
For the moment, Google AI Edge authorizes different models of languages present on Hugging Face and offers three features: “Ask Image”, “Prompt Lab” and “AI Chat”. The first ensures the analysis of an image, the second is designed to synthesize or rewrite text (including computer code), the last consists of a chatbot, in an operation similar to that of Gemini or Chatgpt.
Limitations and execution power
By default, it is possible to download on the four-language language models: Gemma-3n-E2B-IT-Int4, Gemma-3N-E4B-IINT4, Gemma3-1B-IT Q4, and QWEN2.5-1.5B-Partuct Q8. However, we note the possibility of importing others.
When we talk about executing a local model, it is crucial to consider the hardware part, that is, the required power of the components that equip the smartphone on which the language will turn. The response time to a query will therefore strongly depend on the smartphone model used, some devices being faster than others depending on their power.
In the same line, it is useful to approach the point of the weight of a model: it is generally in gigabyte, some easily exceeding 4 GB.
As this article is written, Google Ai Edge is only available on Android. Nevertheless, the firm indicates that it wants to bring its application to iOS very soon.
