---
id: libtpu
version: "0.0.46"
license: Google Cloud Platform Terms of Service
license_treatment: unclear
maintenance: active
---

# libtpu — Google Cloud TPU runtime library.

License: unclear · Maintenance: active · Downloads: 963.6K/mo

## What it is and what it does

libtpu is Google's core runtime library that bridges machine learning frameworks—JAX, PyTorch, and TensorFlow—to Google Cloud TPU hardware. It handles the low-level work of compiling models, managing inter-chip communication, and orchestrating execution on TPU devices. The library also exposes SDK primitives for direct TPU interaction and deployment workflows.

The package is tightly coupled to Google Cloud TPU infrastructure and requires Python 3.11 or newer. It ships as platform-specific wheels for manylinux_2_31 x86_64 only, with no runtime dependencies beyond the Python standard library. Installation is straightforward via pip, but actual use requires access to Google Cloud TPU hardware.

Use it for:

- Running JAX models on Google Cloud TPUs for large-scale machine learning training and inference.
- Deploying PyTorch models to TPU clusters via Google Cloud for distributed training.
- Executing TensorFlow workloads on TPU hardware with automatic compilation and optimization.
- Building custom TPU deployment pipelines using the SDK primitives for direct TPU control.
- Scaling inference workloads across multiple TPU chips with built-in inter-chip communication.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

libtpu is the runtime library that enables JAX, PyTorch, and TensorFlow to run models on Google Cloud TPUs, handling compilation, inter-chip communication, and execution.

Yes, if you are running machine learning workloads on Google Cloud TPUs with JAX, PyTorch, or TensorFlow. The package is actively maintained, has no known vulnerabilities, and is essential for TPU access. No, if you do not have TPU hardware or are not using Google Cloud—the library is platform-specific and will not function without it. Review Google's licensing terms before production use.

## Install

pip install libtpu
uv add libtpu
poetry add libtpu

## Installing libtpu

Before you install: Medium install friction due to platform-specific wheels (manylinux_2_31 x86_64 only); requires Python 3.11 or newer. Package is actively maintained with recent releases and no known vulnerabilities.

License in practice: Licensed under Google Cloud Platform Terms of Service, which is not a standard open-source license. Review Google's terms before deploying in production or redistributing.

Quickstart:

pip install libtpu
import libtpu
# Use with JAX, PyTorch, or TensorFlow on Google Cloud TPU

Requires Google Cloud TPU hardware access and manylinux_2_31 x86_64 platform; Python 3.11 or newer.

Verify before relying:

- Whether libtpu can be installed and used outside of Google Cloud TPU environments or requires TPU hardware.
- Exact API surface and SDK primitives available for direct TPU interaction mentioned in the description.
- Compatibility matrix with specific JAX, PyTorch, and TensorFlow versions beyond the JAX 0.7.1+ note.

## Package facts

- License: Google Cloud Platform Terms of Service (unclear)
- Python support: supports_current
- Install friction: medium
- Maintenance: active
- Downloads: 963.6K/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

tpu runtime library, google cloud tpu support, tpu machine learning framework, tpu compilation and execution, jax pytorch tensorflow tpu, tpu inter-chip communication, cloud tpu sdk, tpu-runtime, google-cloud, ml-acceleration

[View on SkillFed](https://skillfed.io/packages/libtpu) · [View on PyPI](https://pypi.org/project/libtpu/)