Understand this story
At PyTorch Conference North America 2026, Google and Meta open-sourced TorchTPU, a native PyTorch backend for Google TPUs that replaces PyTorch/XLA from 2020. According to an InfoQ AI report, migrating a model only requires changing model.to("cuda") to model.to("tpu"). The official release also open-sources four repositories: TorchTPU, vLLM-TorchTPU, SGLang-TorchTPU, and TorchTitan; Hugging Face models can run directly, with native support from the vLLM and SGLang inference engines. Google's published results include running the Qwen3 235B MoE model on 8192 TPU v7x chips and achieving 28% model FLOPs utilization; the announcement also notes that a one-line change only gets the model running, and reaching peak performance still requires tuning shapes, sharding, and operators. There are currently no subsequent developments.
AI-generated from public reports
Latest development · Report
Google and Meta Open-Source TorchTPU at PyTorch Conference 2026Developments
1 confirmed developments
- Source:InfoQ AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 26
InfoQ AISelected
Google and Meta Open-Source TorchTPU at PyTorch Conference 2026
Google and Meta open-sourced TorchTPU at PyTorch Conference North America 2026. This is a native PyTorch backend for Google TPUs, replacing PyTorch/XLA from 2020; migrating a model only requires changing model.to("cuda") to model.to("tpu"). The official release also open-sources four repositories: TorchTPU, vLLM-TorchTPU, SGLang-TorchTPU, and TorchTitan. Hugging Face models can run directly, with native support from the vLLM and SGLang inference engines. Google's published results include running the Qwen3 235B MoE model on 8192 TPU v7x chips and reaching 28% model FLOPs utilization, but the announcement also states that a one-line change only gets the model running; achieving peak performance still requires adjusting shapes, sharding, and operators.
AI Trends