Gemma 4 2B

Overview

Google's 2B base model in the Gemma 4 family with text and image input, 128k context, and a 1.2GB Q4 footprint that runs on integrated graphics or a Raspberry Pi 5.

When to pick this model

• On-device assistants for laptops, phones, and SBCs
• Multimodal prototypes that can't justify a dedicated GPU
• Long-context summarization at the edge
• Air-gapped or offline scenarios where latency and privacy matter

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	1.2 GB
Q5_K_M	1.4 GB
Q8_0	2.1 GB
FP16 (no quantization)	4 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Strengths

• Runs on integrated GPUs at ~1.2GB VRAM in Q4
• Multimodal text and image input out of the box
• 128k context unusual at this parameter count
• Permissive Gemma license

Limitations

• Reasoning lags behind 4B and larger Gemma variants
• Gated on Hugging Face (click-through access)

Architecture & training

Architecture: Gemma 4 base · 2B dense · multimodal text + image · 128k context

Training: Google Gemma 4 family, 2B multimodal base version, trained for edge/laptop.

Quick start

ollama run gemma4

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.