[digitallyfree]

Theoretically the ARC should work with llama.cpp using OpenCL, but I haven't seen benchmarks or even a confirmation that it works.