AI Potluck
Contribute
Back to Gap Map Infrastructure / ML orchestration

Metaflow

Netflix
open source / Overall score: 3.0

Python framework from Netflix for building and operating machine-learning and AI systems. A flow is a class of steps; decorators attach resources, retries and checkpoints to each step, and @batch or @kubernetes send that step to AWS Batch or a Kubernetes cluster to run. Metaflow tracks every run, its artifacts and its models, and can resume a failed flow from the step that failed.

Outerbounds, founded by Metaflow's creators, sells a managed platform beside it. Verified 2026-09-15 via the repository, its recursive tree and the PyPI download API.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(Netflix/metaflow is the framework you run)
core-gated
ungated(a 1,255-entry recursive tree carries a root LICENSE and R/LICENSE, every other license file under metaflow/_vendor/, and no path matching enterprise, ee, commercial or proprietary)

Apache-2.0 with the whole framework published. Outerbounds sells a managed platform alongside, which the ladder does not read as gating. The 1,255-entry recursive tree carries one governing LICENSE; every other license file is a vendored third-party dependency under metaflow/_vendor/, and there is no enterprise, ee or commercial path.

  • https://github.com/Netflix/metaflow recorded 2026-09-15

    Repository page for Netflix/metaflow: Apache-2.0 license, public and unarchived, described as 'Build, Manage and Deploy AI/ML Systems'. Establishes the license and that the source is public.

  • https://api.github.com/repos/Netflix/metaflow/git/trees/master?recursive=1 recorded 2026-09-15

    Full untruncated recursive tree of the default branch, 1,255 entries. Licenses are the root LICENSE and R/LICENSE; every other license file sits under metaflow/_vendor/. No path matches enterprise, ee, commercial or proprietary. A tree lists paths; it is cited for that and for nothing about the vendor's offerings.

  • https://outerbounds.com/ recorded 2026-09-16

    Outerbounds' homepage, describing its platform as a separate product built around Metaflow and announcing that 'Outerbounds is joining Anaconda.' Establishes a commercial offering beside the open framework rather than a paid edition of it.

Adoption

3 high confidence
3.0

The PyPI download API reports 482,037 downloads of metaflow in the trailing 30 days, inside the 100K-1M band for level 3. PyPI is the product's declared distribution channel, so the level can be recomputed from that feed.

Capability

3 medium confidence
3.0

Admitted on the accelerator limb - @resources(gpu=...) is Metaflow's own decorator, not a pod spec passed through. Band 3: Metaflow decides when a step runs and which queue or cluster it is submitted to, and AWS Batch or the Kubernetes scheduler chooses the machine. Under this dimension it does what flyte does, and the two are banded together.

Verified 2026-09-15