Docker Desktop 4.71.0, released April 27, 2026, introduces a behavioral change that can silently break existing workflows: Docker Model Runner is now disabled by default. Anyone who configured it in a previous version will find the service off after updating, with no explicit error on first launch.
What the New Default Means
In previous versions, Docker Model Runner started automatically with Docker Desktop. Starting with 4.71.0, the service must be manually enabled from Settings > Features in development > Enable Docker Model Runner.
Once enabled, TCP host-side support activates automatically, with no separate configuration needed. The API behavior is unchanged for anyone who re-enables it.
The most likely breakage scenarios are:
- Local CI/CD pipelines that use Model Runner for LLM-based tests: if the script assumes the service is running, it will fail silently or with a connection error.
- Docker Compose stacks with services that depend on the Model Runner endpoint: the dependency won’t resolve until Model Runner is explicitly enabled.
- Environment setup scripts that don’t include an enable step: after an update or a fresh install on a new machine, Model Runner won’t be available without manual intervention.
Docker didn’t explain the reasoning in the changelog. The pattern fits a footprint-reduction decision: Model Runner consumes RAM and CPU even at idle, and the majority of Docker Desktop users don’t use it. Making it opt-in is the straightforward way to reduce resource usage on machines that don’t need it.
Updated Components
The 4.71.0 release includes updated bundled components:
- Docker Engine 29.4.1
- containerd 2.2.3
- Runc 1.3.5
- Docker Compose 5.1.3
- Docker Agent 1.44.0
- Docker Model Runner v1.1.36
RHEL 8 Support Ending
The next Docker Desktop release will require RHEL 9 or RHEL 10 for installation on Red Hat Enterprise Linux. Support for RHEL 8 ends with 4.71.0. Teams running Docker Desktop on RHEL 8 machines in test or build environments should plan the OS upgrade before the next release makes it a hard requirement.
Mac Fix
The 4.71.0 release corrects an issue where error tracking continued transmitting session data after the user had disabled analytics collection. It is a compliance fix rather than an operational one, but relevant for teams operating under strict telemetry policies.
What to Do Before Updating
If Docker Model Runner is part of your workflow, check which scripts or pipelines assume the service is already running and add an explicit enable step or a connection check. The alternative is to manually re-enable it right after updating — three clicks in Settings — but it is worth documenting the requirement before the next time a colleague does a fresh Docker Desktop install and wonders why the model endpoint is not responding.
