How to install TensorFlow with pip

A Python project needs the TensorFlow package in the same environment that runs notebooks, training scripts, or inference code. The official PyPI wheel keeps the install inside a selected Python environment without moving the project to a separate Conda workflow.

Running pip through python -m pip ties the package operation to the interpreter selected by the shell or virtual environment. Confirm the interpreter first, because TensorFlow wheels target a narrower set of CPython versions than many workstation defaults.

The plain tensorflow package is the CPU install path for supported platforms. Use tensorflow[and-cuda] only on Linux or WSL2 systems where the NVIDIA driver already works; native Windows GPU support is limited to legacy TensorFlow releases, native Windows CPU installs still need the Microsoft Visual C++ Redistributable, and current official macOS wheels are CPU-only for Apple Silicon.

Steps to install TensorFlow with pip:

Activate the Python environment that should receive TensorFlow.
```
$ source ~/venvs/tf/bin/activate
(tf) $
```
On Windows, run the equivalent virtual environment activation command or use a matching Python Launcher selector such as py -3.12 -m pip for the remaining pip commands.
Check that the active interpreter is a supported Python version.
```
(tf) $ python --version
Python 3.12.13
```
Current TensorFlow 2.21 wheels cover CPython 3.10 through 3.13 on supported platforms. Switch environments before installing if this command reports Python 3.14 or newer.
Confirm that pip points inside the active environment.
```
(tf) $ python -m pip --version
pip 26.1.2 from /home/user/venvs/tf/lib/python3.12/site-packages/pip (python 3.12)
```
The path should belong to the environment you activated. A system path means the install would modify a different Python package set.

Upgrade pip in the active environment.

(tf) $ python -m pip install --upgrade pip
Requirement already satisfied: pip in /home/user/venvs/tf/lib/python3.12/site-packages (25.0.1)
Collecting pip
##### snipped #####
Successfully installed pip-26.1.2

Install the TensorFlow package from PyPI.
```
(tf) $ python -m pip install tensorflow
Collecting tensorflow
  Downloading tensorflow-2.21.0-cp312-cp312-manylinux_2_27_aarch64.whl.metadata (4.4 kB)
##### snipped #####
Successfully installed tensorflow-2.21.0
```
The wheel tag, dependency versions, and CPU architecture can differ by platform. On supported Linux or WSL2 GPU hosts, install tensorflow[and-cuda] instead after confirming the NVIDIA driver works.
Related: How to enable GPU acceleration in TensorFlow
Confirm that the active environment can import TensorFlow.
```
(tf) $ python -c "import tensorflow as tf; print(tf.__version__)"
2.21.0
```
Related: How to check the TensorFlow version
Run a small tensor operation through the installed runtime.
```
(tf) $ python -c "import tensorflow as tf; print(tf.reduce_sum(tf.ones([1000, 1000])))"
tf.Tensor(1e+06, shape=(), dtype=float32)
```
The tensor result confirms that the package imports and executes a TensorFlow operation in the active environment.