Nvidia's New AI Runs on Your Phone, No Internet Needed
This tiny AI could make your phone assistant faster and smarter—without the cloud.
NVIDIA’s Nemotron-3.5-Bolt-E4B-4GB is a small language model post-trained from Gemma-4-E4B-it. It carries forward Gemma’s conversational capabilities and improves multi-turn tool use, instruction following, and input grounding for fast local voice/device assistants, interactive personas, and embedded/edge interfaces.
Designed for low-memory, low-TTFT operation, it lets agents respond quickly while preserving headroom for applications or adjacent model pipelines. It is ready for commercial or non-commercial use, and a DFlash drafter is available for speculative decoding.
- Nvidia's new AI model is small enough to run on phones and laptops, enabling fast, offline voice assistants.
- It's free for commercial use, so device makers can integrate it without licensing fees.
- The model prioritizes speed and low memory, but may not match the intelligence of larger cloud AIs.
Why It Matters
This could lead to faster, more private AI assistants on your devices, reducing reliance on the cloud.