Infrastructure / Core ML frameworks & libraries

Keras

Keras is a high-level deep learning framework billed as 'Deep Learning for humans'; Keras 3 is multi-backend, running on JAX, TensorFlow, PyTorch, and OpenVINO. It is maintained by the keras-team organization (closely tied to Google) and reports use by nearly three million developers. It provides an accessible API for building and training models while letting users choose the underlying backend.

Verified live 2026-06-22 via primary sources. Apache-2.0 licensed with full public source.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
core-gated: ungated

Apache-2.0 licensed with full public source.

https://github.com/keras-team/keras recorded 2026-06-22
Repo shows Apache-2.0 license and full source

Adoption

5 high confidence

5.0

Re-scored 4->5. 21.53M PyPI downloads in the last 30 days. The map's prevailing usage_volume level-5 floor is >10M/mo (cf. pydantic-ai ~31M, llama-index ~10.18M at level 5); the prior level 4 applied a stricter 5M-50M band used only in one ml_frameworks batch, below map peers of equal volume. Keras is the standard high-level training API across TF/JAX/PyTorch and clears the prevailing floor.

https://pepy.tech/projects/keras recorded 2026-06-25
21,532,407 downloads in last 30 days

Capability

4 high confidence

4.0

Central API layer but sits atop lower-level frameworks rather than defining the compute layer.

https://github.com/keras-team/keras recorded 2026-06-22
README describes multi-backend deep learning framework

Unchanged since 2026-07-30 (last edited, not re-checked)