Introducing @huggingface/kernels: 200+ WebGPU Kernels for Local AI
Hugging Face released @huggingface/kernels, alongside an initial collection of 207 versioned WebGPU kernels on the Hub (webgpu-kernels), and benchmarked them against ORT WebGPU on an Apple M4 GPU where they were 2.57× faster by geometric mean and 1.90× faster at the median across 809 matching-and-timed test cases.
Sep 1, 2026 ·Hugging Face

















