TensorBoard
TensorFlow's open-source dashboard for tracking training metrics and visualising models and data.
Updated · AppBreakup editorial · Windows, macOS, Linux, Self-hosted
Is TensorBoard free?
Yes, free and open source
- License:
- Apache-2.0
Checked on TensorBoard's site Oct 30, 2026
Switch from TensorBoard to
All 3
NetronOpen source
Model ExplorerOpen source
Weights & BiasesStarting price: From $60 a month (Pro), billed monthly
- Verified facts
- 4 of 4
- Alternatives ranked
- 3
- Approved reviews
- 0
- Visits sent from here
- 0
About TensorBoard
TensorBoard is the visualisation toolkit that ships with TensorFlow and also works with PyTorch through its logging writer. You point it at a folder of training logs and it opens a local web dashboard with charts of loss, accuracy and other metrics over time, the model graph, histograms of weights and biases, embedding projections, and sample images, text and audio. A profiler helps find slow steps in training. It is free and Apache-licensed and runs wherever Python runs, viewed in your browser. It is aimed at people training models who want to watch a run and compare runs side by side. It is less suited to teams who need shared experiment history, user accounts or a model registry, since it reads local log files rather than storing runs in a server.
Best for: People training TensorFlow or PyTorch models who want live charts of each run.
Pros and cons
What works
- Metrics, graph, histograms, embeddings and profiler in one dashboard
- Free, open source and already included with TensorFlow
- Works with PyTorch logs too
Why people leave
- Reads local log folders; no built-in sharing, accounts or run database
- Graph view is tied to TensorFlow-style logs rather than any model file
TensorBoard vs Netron, Model Explorer and Weights & Biases
From each app's listing and the prices verified on its own site. A dash means we haven't verified it yet.
| Attribute | TensorBoard | Netron | Model Explorer | Weights & Biases |
|---|---|---|---|---|
| Pricing | Open source | Open source | Open source | Free plan + paid |
| Verified price | — | — | — | From $60 a month (Pro), billed monthly |
| Free plan | Free app | Free app | Free app | Yes, $0 a month |
| Open source | Yes | Yes | Yes | No |
| Self-hostable | Yes | No | No | No |
| Platforms | Windows, macOS, Linux, Self-hosted | Web, Windows, macOS, Linux | Windows, macOS, Linux | Web, iOS |
Top TensorBoard alternatives
See all 3- 1Netron64.8
Opens almost any saved model file directly when all you need is to inspect its structure.
- 2Model Explorer61.8
A dedicated graph viewer that copes better with very large models through collapsible layers.
- 3Weights & Biases42.3
Shares and compares runs across a team in a hosted dashboard.
Why people left TensorBoard
In reviewers' own words, from reviews of the apps they moved to. These are their opinions, not ours.
No switching stories yet. If you moved away from TensorBoard, review the app you switched to and tell us why.
Reviews
No approved reviews yet. How we verify reviewers
Write a review of TensorBoard
TensorBoard: common questions
What is TensorBoard?
TensorBoard: TensorFlow's open-source dashboard for tracking training metrics and visualising models and data. Best for people training TensorFlow or PyTorch models who want live charts of each run. Available for Windows, macOS and Linux, and you can host it yourself.
Is TensorBoard free?
Yes. TensorBoard is free, open-source software that you can host yourself.
Is TensorBoard open source?
Yes. The source code is on GitHub at tensorflow/tensorboard. You can also run it on your own server.
What platforms does TensorBoard run on?
TensorBoard is available for Windows, macOS, Linux and Self-hosted.
What are the best alternatives to TensorBoard?
The highest-ranked TensorBoard alternatives right now are Netron, Model Explorer and Weights & Biases. See the full ranked list for pricing, platforms and verified facts.
TensorBoard vs Netron: what's the difference?
Both TensorBoard and Netron are free and open source. Only Netron runs on Web.