Comment by pixl97

They are busy running training/inference on them and haven't benchmarked them?