# Serve the model

Modern LLM applications require scalable, high‑performance serving engines that can efficiently host large models and deliver low‑latency inference.
The Cloud AI SDK supports the following serving backends to give you flexibility in how models are deployed, optimized, and integrated into production workflows. This section provides a unified view of these serving options and helps you select the right runtime for your workload.

- [vLLM](https://docs.qualcomm.com/doc/80-99100-3/topic/index_vLLM-Serving.html)
- [Triton Inference Server](https://docs.qualcomm.com/doc/80-99100-3/topic/index_Triton-Serving.html)
- [Text Generation Inference (TGI)](https://docs.qualcomm.com/doc/80-99100-3/topic/index_TGI-Serving.html)

After successfully serving your model using vLLM, Triton, or TGI, you can further optimize performance or implement specialized logic using [advanced model techniques](https://docs.qualcomm.com/doc/80-99100-3/topic/index_Advanced-model-techniques.html#advanced-model-techniques).

Last Published: Aug 25, 2026

[Previous Topic
Deploy the model with Kubernetes](https://docs.qualcomm.com/bundle/publicresource/80-99100-3/topics/index_Kubernetes.md) [Next Topic
vLLM](https://docs.qualcomm.com/bundle/publicresource/80-99100-3/topics/index_vLLM-Serving.md)