NVIDIA Supports Google DeepMind’s Gemma 3n on Jetson and RTX

Zach Anderson
Jun 26, 2025 12:16

NVIDIA announces support for Google DeepMind’s Gemma 3n on Jetson and RTX platforms, enhancing AI capabilities across devices with optimized models.

NVIDIA has announced the support for Google DeepMind’s Gemma 3n models on its RTX and Jetson platforms, a significant step in advancing artificial intelligence capabilities, according to NVIDIA’s official blog. This development follows the preview of Gemma by Google DeepMind at Google I/O, showcasing two new models optimized for multi-modal on-device deployment.

Gemma 3n’s Enhanced Capabilities

The Gemma 3n models now incorporate audio, alongside existing text and vision capabilities introduced in version 3.5. The models integrate renowned research components, such as the Universal Speech Model for audio, MobileNet v4 for vision, and MatFormer for text processing. A notable advancement is the Per-Lay Embeddings innovation, which significantly reduces RAM usage, allowing higher quality models to operate within resource-constrained environments.

Jetson’s Role in Robotics and Edge AI

NVIDIA Jetson devices are set to benefit from the lightweight architecture and dynamic memory usage of the Gemma models, making them ideal for next-generation robotics and other edge applications. Developers have the opportunity to engage in the Gemma 3n Impact Challenge on Kaggle, aiming to leverage this technology for positive global changes in various fields, including accessibility, education, and healthcare.

Utilizing NVIDIA RTX for AI Enthusiasts

For AI enthusiasts and developers, NVIDIA RTX AI PCs offer a platform to deploy Gemma 3n models with ease, utilizing Ollama for efficient deployment. The models can be integrated into applications such as AnythingLLM and LM Studio, with NVIDIA collaborating with Ollama to optimize performance on RTX GPUs.

Customization with NVIDIA NeMo Framework

Developers can further customize Gemma 3n models using the open-source NVIDIA NeMo Framework, accessible via Hugging Face. This comprehensive framework supports post-training fine-tuning with enterprise-specific data, enhancing model accuracy through techniques like Low-Rank Adaptation and Parameter-Efficient Fine-Tuning.

With NVIDIA’s commitment to open-source projects, the company continues to foster AI transparency and collaboration, promoting the development of safe and resilient AI models.

For more detailed information on NVIDIA’s support for Google DeepMind’s Gemma 3n, visit their official blog.

Image source: Shutterstock