Model5d ago
Gemma 4: 4B Multimodal Model Runs On Device
Gemma 4, a 4B‑parameter multimodal model, runs on a single 8GB GPU, eliminating the need for cloud inference. Hugging Face’s new release supports image, text, and audio inputs, delivering real‑time responses in under 200 ms on edge devices. Developers can integrate it into mobile apps without additional server costs. The model demonstrates that high‑quality multimodal AI can be deployed locally.