OSS TanbouOverview
Keras 3 provides a high-level model-development interface across JAX, TensorFlow, and PyTorch backends. OpenVINO is inference-only, while use cases include vision, language, audio, time series, and recommendation models.
Features and best fit
Based on official documentation; not hands-on tested · Content checked:
Choose a backend and scale through one model API
It keeps a high-level development experience while selecting JAX, TensorFlow, or PyTorch and scaling training from laptops to GPU or TPU clusters. The pip package is keras, and the chosen backend package is installed separately. OpenVINO is limited to inference.
Sources: [1]
For teams balancing readable model code with backend choice
It suits teams testing vision, NLP, audio, forecasting, and recommendation through a common interface while choosing a backend around hardware and existing assets. Validate numerical differences, speed, distributed training, serialization, and custom-layer compatibility on the target model.
Sources: [1]
Do not confuse Keras 3, tf-keras, and backend initialization
Keras 3 is published as keras, while Keras 2 remains available as tf-keras. Select the backend through an environment variable or config file before importing keras; it cannot be changed afterward. The README recommends WSL2 on Windows, and GPU environments must align backend-specific CUDA dependencies with the driver. Keras uses Apache-2.0.
Sources: [1]
Official sources
- [1]Keras 3 README(2026-09-13)
Supplemental curator note
Selected for pairing a human-centered tagline and concise API with new version- and backend-specific intent created by the Keras 3 transition.
Growth
Growth trends · Last 30 days
64,320 Stars
Trend data is still being collected.
Development activity
Last 90 days · weekly
- Commits (last 30 days)
- 104
- Open PRs
- 62
Development activity is still being collected.
Built with
Categories and tags
Categories
GitHub data
GitHub dataView detailed GitHub data
GitHub Topics
- deep-learning
- tensorflow
- neural-networks
- machine-learning
- data-science
- python
- jax
- pytorch
- Stars
- 64,320
- Forks
- 19,788
- Watchers
- 1,853
- Open issues
- 129
- Primary language
- Python
- License
- Apache-2.0
- Repository last updated
- Sep 12, 2026
Explore next
- TensorFlow200,054 Stars
An end-to-end machine-learning platform connecting research, training, evaluation, visualization, and deployment.
C++ - Ollama180,804 Stars
A runtime for downloading and running open models on a computer or server through a CLI and local API.
Go - Transformers165,353 Stars
A Python framework for inference and training with pretrained text, vision, audio, and multimodal models.
Python - AUTOMATIC1111 Stable Diffusion WebUI164,915 Stars
A Gradio UI for Stable Diffusion image generation, editing, model selection, extensions, and APIs.
Python - Dify155,591 Stars
A visual LLM application platform combining AI workflows, RAG, agents, model management, and observability.
TypeScript
Report incorrect information
Tell us if any listing information is incorrect or outdated.