Gemma 3n — local AI model by Google DeepMind
Written by Jakub Rusinowski · Last updated
Google's mobile-first multimodal model family. Uses the novel MatFormer nested architecture — a single model file contains multiple sub-models (E2B/E4B) that can run at different sizes. Processes text, images, audio, and video. Runs on phones without internet.
Variants
The smallest Gemma 3n variant needs about 4 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Gemma 3n E2B → 2B | ~4.1 GB | 32,768 | ollama run gemma3n:e2b | |
| Gemma 3n E4B → 4B | ~5.5 GB | 32,768 | ollama run gemma3n:e4b |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Gemma 3n locally
Install Ollama, then pull the tag.
ollama run gemma3n:e2bPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Applies to: Gemma 3n E2B, Gemma 3n E4BRecommended GPU
The cheapest catalogued GPU that runs Gemma 3n locally (min 4 GB VRAM) is the Intel Arc B570 (10 GB).
Can I run Gemma 3n on my GPU?
Gemma 3n — frequently asked questions
How much VRAM does Gemma 3n need?
Gemma 3n needs about 4 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Gemma 3n E2B (4 GB, Q4_K_M); Gemma 3n E4B (6 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Gemma 3n on an RTX 4090 (24 GB)?
Yes — Gemma 3n runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Gemma 3n?
Q4_K_M is the best balance of quality and VRAM for Gemma 3n in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Gemma 3n with Ollama?
Install Ollama, then run: ollama run gemma3n:e2b. This downloads Gemma 3n and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.