Skip to main content
Local LLM Guide

Model

Gemma 3n E4B Instruct

Start with a supported compressed build and short context. 16GB RAM estimate. Best use: mobile and laptop local assistants

8B total / 4B activeMultimodal

Best use

mobile and laptop local assistants · multimodal extraction

Memory planning

16GB RAM estimate

Expected pace

Depends on your hardware, app, and conversation length

License

Gemma Terms of Use; read the model’s terms before use

Who should use this model?

  • mobile and laptop local assistants
  • multimodal extraction
  • speech and audio understanding
  • efficient on-device chat

Choose something else when

  • the local runner needs mature fully supported multimodal GGUF support
  • a precise public training-token count is required

What hardware makes sense?

The download is only part of the memory needed. Leave room for the app, the input it processes, and other open programs. For chat models, context includes the instructions, conversation, and answer. Longer context needs more memory.

CPU only
A possible starting point with a compressed model and short conversation; check the exact file size.
8GB system memory
A possible starting point with a compressed model and short conversation; check the exact file size.
16GB system memory
A possible starting point with a compressed model and short conversation; check the exact file size.
32GB system memory
A possible starting point with a compressed model and short conversation; check the exact file size.
8GB GPU memory
A possible starting point with a compressed model and short conversation; check the exact file size.
12GB GPU memory
A possible starting point with a compressed model and short conversation; check the exact file size.
24GB GPU memory
A possible starting point with a compressed model and short conversation; check the exact file size.
Apple unified memory
A possible starting point with a compressed model and short conversation; check the exact file size.

Model facts that change the decision

Parameters
8B total / 4B active
Input
Text · Image · Audio · Video
Published context limit
32,768 tokens
Model format
safetensors · GGUF · Ollama · Transformers · MLX · llama.cpp · Google AI Edge
Suggested compression
Q4_K_M for balanced local use; Q5_K_M when memory allows; use text-only GGUF if the target runner lacks multimodal support.
Typical Q4 file
4.24GB before runtime and context overhead
Languages
English · 140+ spoken languages

Using image input

This model needs an app or library that supports its image-processing components. Follow the exact model’s instructions for preparing image inputs. Some GGUF downloads also need a matching projector file; loading the text model alone does not establish image support.

First setup

Ollama

ollama run gemma3n:e4b

LM Studio: Search for bartowski/google_gemma-3n-E4B-it-GGUF or lmstudio-community/gemma-3n-E4B-it-text-GGUF; verify whether the selected build supports multimodal input.

llama.cpp

Use hf.co/bartowski/google_gemma-3n-E4B-it-GGUF:Q4_K_M for text inference; verify multimodal/audio support in the llama.cpp build before enabling non-text inputs.

Conditions that change the answer

  • Selected E4B rather than E2B because E4B contains the E2B sub-model and is the better-supported higher-capability local choice while remaining small-device oriented.
  • Gemma 3n uses effective-parameter terminology: E4B is an effective 4B configuration inside a larger raw-parameter model.
  • Ollama's catalog page currently lists Text input for gemma3n tags, while Google's official documentation describes text, image, audio, and video inputs; treat local multimodal support as runner-dependent.
  • Allow memory for the model file, the app, and your conversation.

Read the controlling license text →