This stack uses Ollama with Gemma 4 QAT to run a 12B model on a 10GB VRAM laptop GPU. The latest Gemma 4 QAT checkpoints reduce memory usage and enable compact local inference.

What you get

• Local Gemma 4 12B inference on 10GB VRAM hardware
• QAT compression that fits the model into ~6.7 GB VRAM
• A laptop-friendly private AI stack for writing, notes, and prompts

Prerequisites

• A laptop with at least 10 GB VRAM, such as RX 6700 series
• Latest GPU drivers and Vulkan support
• Ollama installed locally
• Enough disk space for the model cache (~40 GB)

Setup

brew install ollama
ollama pull gemma-4:12b --quantization qat
ollama serve
ollama ps

If ollama ps shows the model and GPU usage, your stack is ready.

Use it

• Personal writing with faster local completion
• Private research without sending queries to the cloud
• Compact local AI demos on 10GB-class laptops

Troubleshooting

• Model won’t load: verify Vulkan and free VRAM.
• Ollama falls back to CPU: check ollama ps and update drivers.
• Slow inference: close background apps and use the QAT model.

Originally published on https://everylocalai.com/stack/gemma-4-qat-10gb-laptop