Gemma 3 12B

Overview

The 12B sweet spot of Google's Gemma 3 line — multimodal, 128K context, and 140 languages. Fits on a single consumer GPU with room for batching.

When to pick this model

• You want strong multimodal performance on a 16GB or 24GB GPU
• You need long-context summarization or document Q&A with vision
• You're shipping a product covering many languages
• You want one general-purpose Gemma without going to 27B

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	7 GB
Q5_K_M	9 GB
Q8_0	13 GB
FP16 (no quantization)	24 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

• Sweet spot for multimodal performance vs hardware cost
• 128K context window
• 140 language coverage
• Strong general-purpose default

Limitations

• Gemma License rather than Apache
• At least 9GB RAM required for Ollama deployment
• No dedicated thinking mode

Architecture & training

Architecture: Dense VLM · sliding-window attention · multimodal

Training: 12T tokens.

Quick start

ollama run gemma3:12b

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