Docker Desktop LLM support with NVIDIA
Daniel Nashed – 5 May 2025 10:48:17
After my first and simple test, I updated Docker Desktop on my lab machine.
There is a GPU option in the experimental settings showing up when you have a matching GPU.
Once enabled it, the following command loads the model into the NVIDIA GPU
docker model run ai/qwen3:0.6B-Q4_0
To check details I ran a Docker container with Ubuntu also using the GPU.
Beside nvidia-smi to check the card driver version I installed "nvtop" (which is part of Ubuntu) to see the GPU performance.
docker run --gpus all -it --rm ubuntu bash
Looking a bit deeper into the installed binaries you can indeed see that Docker also leverages the llama.cpp project (https://github.com/ggml-org/llama.cpp)
Interestingly the GPU option mentions additional software will be downloaded. But so far I only found those files:
Directory of C:\Program Files\Docker\Docker\resources\model-runner\bin
05/05/2025 12:32
05/05/2025 12:32
05/05/2025 12:31 71 com.docker.llama-server.digest
05/05/2025 12:31 1.838.320 com.docker.llama-server.exe
05/05/2025 12:31 44.784 com.docker.nv-gpu-info.exe
05/05/2025 12:31 481.520 ggml-base.dll
05/05/2025 12:31 492.272 ggml-cpu.dll
05/05/2025 12:31 65.776 ggml.dll
05/05/2025 12:31 1.212.144 llama.dll
7 File(s) 4.134.887 bytes
- Comments [0]