# Hugging Face releases TorchTPU for native PyTorch on Google Cloud TPUs

> The new TorchTPU package lets PyTorch code run directly on Google Cloud TPUs, removing the need for torch_xla or Optimum TPU.

Oossa · 2026-10-20 · https://oossa.com/en/hugging-face-releases-torchtpu-for-native-pytorch-on-google-cloud-tpus

Hugging Face announced TorchTPU, a native PyTorch backend that treats TPUs like any other device. Install it with a single pip command and you can load Transformers or Diffusers models on a TPU by setting device_map="tpu". The package includes TPU‑tuned Pallas kernels, distributed training support, and torch.compile integration for extra speed. Optimum TPU is no longer required for Hugging Face models on TPUs.

## The facts

- Announcement date: 2026-10-20 ("Tue Oct 20 2026")
- Package install command: pip install torch-tpu --extra-index-url https://download.pytorch.org/whl/cpu

## Why it matters

PyTorch developers can now use Google Cloud TPUs without extra frameworks, simplifying large‑scale training and inference.

## Sources & references

1. [Hugging Face ❤️ TorchTPU: TPUs, Natively](https://huggingface.co/blog/hf-torchtpu) – Hugging Face, 2026-10-20
2. [Introducing TorchTPU](https://pytorch.org/blog/introducing-torchtpu/) – PyTorch
3. [Google unveils TorchTPU in a push to break Nvidia’s stranglehold on AI compute](https://siliconangle.com/2026/10/20/google-unveils-torchtpu-in-a-push-to-break-nvidias-stranglehold-on-ai-compute/) – SiliconANGLE, 2026-10-21

Last updated: 2026-10-21