Back to Gap Map Infrastructure / ML orchestration

ClearML

source available / Overall score: 3.0

End-to-end machine-learning platform covering experiment management, data management, pipelines and remote execution. Tasks are enqueued and a ClearML Agent running on a worker machine pulls them, reproduces the original environment in a container and runs the code. The Python SDK is Apache-2.0; the server the platform runs on is licensed under the Server Side Public License.

The product spans three repositories - clearml the SDK, clearml-server and clearml-agent - and the openness score follows the most restrictive license among them.

Openness

2 high confidence

2.0

license: Apache-2.0(the clearml SDK repository)+SSPL-1.0(clearml/clearml-server, the server the platform runs on, under a ClearML Inc copyright)
source: public(the SDK, the agent and the server are all published)

ClearML publishes the full source of its SDK, agent and server, but the server component, clearml-server, carries the Server Side Public License, which forbids running it as a competing hosted service. That license outweighs the Apache-2.0 SDK, so the compound reads the same as if only part of the source were public: readable, not freely runnable.

https://github.com/clearml/clearml recorded 2026-09-15
Repository page for clearml/clearml: Apache-2.0 license, public and unarchived. Establishes the SDK's license and that the source is public.
https://raw.githubusercontent.com/clearml/clearml-server/master/LICENSE recorded 2026-09-15
The clearml-server LICENSE body: 'Server Side Public License, VERSION 1, OCTOBER 16, 2018, Copyright (c) 2025 ClearML Inc.' Establishes the license of the component you actually run.

Adoption

3 high confidence

3.0

Adoption is measured by PyPI downloads of the clearml SDK, the package most users install to run the platform.

https://pypistats.org/api/packages/clearml/recent recorded 2026-09-15
Trailing-window download counts for the declared PyPI artifact: 625,991 downloads in the trailing month.

Capability

3 medium confidence

3.0

ClearML assigns a task to a queue and, with its enterprise dynamic GPU allocation, sets how many GPUs an agent on that queue may use, but the ClearML Agent itself decides which machine polls and runs the task. That keeps ClearML level with MLRun rather than above it.

https://clear.ml/docs/latest/docs/clearml_agent/ recorded 2026-09-15
ClearML Agent documentation describing the execution flow: 'Enqueue a task for execution on the queue. The agent pulls the task from the queue. The agent launches a container in which to run the task's code.'
https://clear.ml/docs/latest/docs/clearml_agent/clearml_agent_dynamic_gpus/ recorded 2026-09-15
ClearML's dynamic GPU allocation documentation, marked an Enterprise feature, describing allocation of GPUs to queues with clearml-agent daemon --dynamic-gpus --gpus 0-2 --queue dual_gpus=2.

Verified 2026-09-15