Docker Inc. has released Docker Desktop 4.40, packing significant improvements for AI workloads and cross-platform developer experience.
Docker Model Runner
The headline feature is the new Docker Model Runner: a built-in service for running quantized LLMs locally with a single docker model run llama3.1:8b command. Models are pulled as OCI artifacts, cached across projects, and exposed via an OpenAI-compatible API on localhost:8080.
WSL2 File System 40% Faster
A rewritten 9P file-sharing driver reduces bind-mount overhead dramatically on Windows. Benchmarks show npm install on mounted source trees completing in roughly 60% of previous time. The legacy driver remains available via settings flag for teams needing it.
Compose Bridge for Kubernetes
Compose Bridge generates production-ready Helm charts and raw Kubernetes manifests from a docker-compose.yml, complete with ConfigMaps, Secrets, and NetworkPolicies. Ideal for moving local dev environments to staging clusters with minimal friction.
The update is rolling out through auto-update and is available for Mac, Windows, and Linux.
