AMD or Intel GPU reports unsupported operations
Local inference is built around CUDA, so non-NVIDIA GPUs need a different checkpoint.
How to fix it
- 01In the model list, pick a GGUF or ONNX variant — these run on CPU and on AMD/Intel accelerations.
- 02For AMD Radeon on Windows, use the CPU fallback build if the ONNX provider fails to load.
- 03Update your GPU driver to the latest version from the vendor, not from Windows Update.
- 04If you are on an older Integrated Intel generation, expect CPU-speed generation only.