The image in this gist contains a workflow to use the LCM lora that lets you get consistent images with a low amount of steps.
The one for SDXL used in the workflow can be downloaded from: https://huggingface.co/latent-consistency/lcm-lora-sdxl/blob/main/pytorch_lora_weights.safetensors
Where to download ModelSamplingDiscrete