← Back to the wire

Hugging Face ❤️ TorchTPU: TPUs, Natively

AnnouncementProductOct 20, 2026

Hugging Face announced TorchTPU, a native PyTorch backend that lets Transformers, Diffusers, and Kernels run directly on Google's TPUs. Users can load Hub models with device_map="tpu" without Optimum TPU or torch_xla. TPUs, which Google uses to train and serve Gemini, are available on Google Cloud. Hugging Face reports 96.5% of Transformers tests pass on TPU, with 34 issues filed and 19 already fixed.

Receipt № 25021 source · awaiting confirmation ◐

Evidence

1source· awaiting independent confirmation

No score is assigned. Sources and their independence are shown in the citation chain below.

Citation chain · 1 source

01high
huggingface.coOfficialhttps://huggingface.co/blog/hf-torchtpuThis is Hugging Face's own announcement, explaining that TorchTPU makes TPU a native PyTorch device and that Optimum TPU is no longer needed to run its models on Google's accelerators.
PRIMARY
GoogleCompanyGeminiModelHugging FaceCompany
Canonical: https://huggingface.co/blog/hf-torchtpu