Back to Gap Map Model components / Inference code

Docker Model Runner

Docker
open source / Overall score: 2.0

Docker Model Runner pulls, runs and serves AI models with Docker's tooling, from Docker Hub, any OCI registry or Hugging Face. It serves models over OpenAI- and Ollama-compatible APIs, uses llama.cpp by default with vLLM and Diffusers as alternative engines, and packages GGUF and Safetensors files as OCI artifacts. It ships inside Docker Desktop and Docker Engine, and as a standalone dmr binary.

Docker Desktop, which bundles Model Runner on macOS and Windows, is proprietary, but the runner, its CLI and the standalone binary are all in the Apache-2.0 repository. The 2026-09-26 model_hubs sweep routed it to a runtime category. It is filed here rather than in deployment, beside Ollama, because deployment bands capability on isolation strength and this is a model server.

Openness

5 high confidence
5.0
license
Apache-2.0 (OSI)
source
public
core features withheld
no

Docker Model Runner is Apache-2.0 licensed, and the repository holds the inference daemon, the model CLI and the standalone dmr binary, which runs without Docker installed. Docker Desktop, one of the ways it ships, is proprietary, but no Model Runner feature is held back for it.

Adoption

1 low confidence
1.0

GitHub stars are the only signal used, and a star is not a use. Most people get Model Runner inside Docker Desktop, which publishes no count, and Docker has published no user figure. The docker/model-runner image is only the Linux Docker Engine channel, and its pull count is a lifetime total that also counts updates, so it is kept as context rather than banded.

Capability

3 medium confidence
3.0

Docker Model Runner sits two tiers below vLLM, level with Lemonade and llamafile among the local runners. It can hand a model to vLLM, but it adds no parallelism or multi-node serving of its own; what it adds is distributing models as OCI artifacts through the Docker toolchain.

  • https://docs.docker.com/ai/model-runner/ recorded 2026-11-12

    The docs list "Serve models on OpenAI and Ollama-compatible APIs", "Support for llama.cpp, vLLM, and Diffusers inference engines (vLLM and Diffusers on Linux with NVIDIA GPUs)", "Package GGUF and Safetensors files as OCI Artifacts", and say Docker Engine "Supports CPU, NVIDIA (CUDA), AMD (ROCm), and Vulkan backends".

Verified 2026-11-12