PyTorch 2.14 Ships Faster Apple Silicon and 100x GPU Speedups

PyTorch 2.14 lands with a new CUTLASS-based GEMM backend, in-place fault tolerance for distributed training, and native linear algebra on Apple Silicon.

Source

PyTorch

Published

Sep 11, 2026, 5:51 PM

Author

AlphaSignal Newsroom

Read

1 min read

PyTorch 2.14 lands with a new CUTLASS-based GEMM backend, in-place fault tolerance for distributed training, and native linear algebra on Apple Silicon.

Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.

Read original report

Next reads

Google AI · news

Google Sends Trillium TPUs to Orbit to Power AI Beyond Earth's Grid

PyTorch · news

PyTorch 2.14 Ships Faster Apple Silicon, Rebuilt Distributed Training and NVGEMM

AlphaSignal · paper

Next reads

Google Sends Trillium TPUs to Orbit to Power AI Beyond Earth's Grid

PyTorch 2.14 Ships Faster Apple Silicon, Rebuilt Distributed Training and NVGEMM

OT-RED Merges Distributed AI Models Across Hospitals With One Upload