New endpoint

What do you want to run?

Choose a prepared model, or bring your own: a Hugging Face link or your own weights.

Any Hugging Face model Paste a link or repo id, or describe the task.

Check

Your own weights A fine-tune or a LoRA adapter, uploaded from the folder save_pretrained wrote.

Model folder safetensors weights with config and tokenizer files, or adapter_config.json with adapter_model.safetensors

Upload and check

Each file's SHA-256 is computed here and pinned: storage refuses any other bytes, and the GPU machine checks every file again before it starts. The check reads the configs and weight headers and refuses anything that would not load, before any price.

Custom container, engine flags or your own cloud account → Advanced setup

What's it for?

Pick the hardware

Choose from the available hardware that fits your model and workload.