Gemma 3n: Google Ai arrives directly on mobile devices | Is gpu a graphics card | Cpu hardware list | Computer hardware parts | Turtles AI

Gemma 3n is the new Google Open Source model, designed to offer advanced advances on mobile devices. Thanks to innovations such as Per-Layer Embedadings, it guarantees high performance with reduced memory consumption, supporting multimodal inputs and operating in offline mode for greater privacy.

Key points:

• Efficiency on mobile devices: thanks to the per-layer embeddings, Gemma 3N significantly reduces the use of the RAM, allowing the execution of complex models on devices with limited resources.
• MULIMODAL SUPPORT: The model manages text inputs, images, audio and video, offering a thorough understanding and advanced interactions.
• Dynamic flexibility: with the "Many-in-1" functionality, Gemma 3n allows the creation of submissive submissions to specific needs, balanced quality and latency.
• Privacy and offline methods: local execution guarantees the protection of user data, allowing operation even without internet connection.

Google recently presented Gemma 3n, a model of AI Open Source designed to operate directly on mobile devices such as smartphones, tablets and laptops. This model represents a significant step towards the accessibility of AI, combining computational power and energy efficiency.

One of the key innovations of Gemma 3n is the use of the per-layer embeddings (PLE), developed by Google Deepmind. This technique allows a significant reduction in the use of RAM, allowing the execution of models with 5 and 8 billion parameters on mobile devices, with a memory consumption comparable to 2 and 4 billion models. In practice, Gemma 3n can operate with a dynamic memory footprint of only 2 GB and 3 GB, making it ideal for on-decision applications.

The model supports multimodal inputs, including text, images, audio and video, offering a thorough understanding and the ability to manage complex interactions. For example, it is able to perform vocal transcriptions in real time and translations from speech to text, all by keeping the user’s data on the device to guarantee privacy.

Gemma 3n also introduces the "Many-in-1" function, which allows the creation of nested subomodels within the main model. This flexibility allows developers to dynamically balance performance and quality according to specific needs, without the need to host separate models. The "Mix’n’mch" function further facilitates the creation of undermodes optimized for specific use cases, offering an ideal compromise between quality and latency.

For developers interested in exploring Gemma 3n, Google offers two ways of access:

-Google to study: a cloud -based platform that allows you to test Gemma 3n directly from the browser, without the need for configuration.

-Google Ai Edge: a suite of tools and bookstores to integrate 3N gem locally on devices, allowing the use of its comprehensive and generation of text and images functionality.

With the launch of Gemma 3n, Google continues to promote a more accessible and respectful AI of privacy, opening new possibilities for intelligent and interactive applications on mobile devices.