Intel Extension for PyTorch (IPEX) is archived and unsupported as of March 2026
[major] framework
Fields
id
gotchas/xpu-ipex-archived-no-supportClass
framework
Severity
major
Concepts
ipexpytorchoneapi
Symptom
The intel-extension-for-pytorch GitHub repository was archived by the owner on Mar 30, 2026. It is now read-only. Intel will not provide or guarantee development, support, maintenance, bug fixes, new releases or updates. Patches are no longer accepted. The project has known security issues. IPEX 2.8 was the last quarterly release.
Root cause
Intel upstreamed most IPEX features and optimizations into PyTorch itself, making the separate extension redundant. Intel now recommends using PyTorch directly for Intel CPU and GPU platforms.
Workaround
Use PyTorch directly instead of IPEX. Most IPEX features (fused attention, INT8 quantization, memory format optimization) are now available in upstream PyTorch. For oneDNN-based optimization, use the oneDNN Graph API directly or through llama.cpp's SYCL backend.
notes
This is a significant change for anyone who built their Intel GPU inference stack on IPEX. The archived status means no security patches, no bug fixes, and no new features. Anyone still using IPEX should migrate to PyTorch directly. The oneDNN library (separate from IPEX) is NOT archived and remains the primary optimization library for Intel GPUs.
Notes
This is a significant change for anyone who built their Intel GPU inference stack on IPEX. The archived status means no security patches, no bug fixes, and no new features. Anyone still using IPEX should migrate to PyTorch directly. The oneDNN library (separate from IPEX) is NOT archived and remains the primary optimization library for Intel GPUs.