Can I deploy a model that isn't in the model library?

Yes — that's most of what dedicated deployments are for. You can deploy:

• Any Hugging Face model — walkthrough

• Custom models packaged with Truss (your code, weights, and dependencies) — model development docs

• LLMs via the engine builder for optimized TensorRT-LLM serving — engine docs

• Models served with vLLM, SGLang, or your own Docker image — see examples

The model library is a convenience, not a boundary.