Draw something on our GPU.
One L40S of ours has Qwen-Image-2.1 Turbo loaded and waiting. Type a prompt and a picture comes back in about five seconds. Two are on us, no account needed; after that it's five cents a picture on your account.
Looking for the box…
Your picture lands here, about five seconds after you send the prompt.
What you just used
- One box, one command
The box drawing your picture is an ordinary session on our fleet, started with one job: a fixed image with the weights baked in, and a script that loads the model once and serves prompts from a folder. Nothing about it is special to us; the same command runs your model, your script, your data.
super run . --script serve.sh --sku gpu_1x_l40s- Watch one boot
- With an account you can run the same demo on a box of your own and watch it find a GPU, boot, download the model and draw, with the cost ticking beside it. Boot a box of your own.