These models are available on the free HF Inference API tier. First call may take 30s while the model loads.