[Doc] Add MatrixHub as a model loading source (#50492)
Signed-off-by: yiting.jiang <yiting.jiang@daocloud.io>
Yiting Jiang committed
Aug 17, 2026 at 10:32 UTC
502af5ed007da93c2a7747b31d0dc0e6110e17a3
1 file changed
+15
docs/models/supported_models.md
+15
@@ -289,6 +289,21 @@ os.environ["http_proxy"] = "http://your.proxy.server:port"
289
os.environ["https_proxy"] = "http://your.proxy.server:port"
290
```
291
292
+### MatrixHub
293
+
294
+[MatrixHub](https://github.com/matrixhub-ai/matrixhub) is a self-hosted model registry and distribution layer that caches models from upstream hubs and serves them over a Hugging Face-compatible API inside your own network.
295
+
296
+Since the API is Hugging Face-compatible, you only need to point `HF_ENDPOINT` at your MatrixHub instance:
297
+
298
+```shell
299
+export HF_ENDPOINT="http://<your-matrixhub-address>"
300
+vllm serve Qwen/Qwen3-0.6B
301
+```
302
+
303
+vLLM then downloads model weights from MatrixHub over the internal network instead of the public Hugging Face Hub, which is useful for air-gapped clusters and for avoiding repeated downloads across nodes.
304
+
305
+See the [MatrixHub guide for vLLM](https://matrixhub.ai/docs/guides/use-with-vllm/) for an end-to-end walkthrough, including Docker and Kubernetes deployment examples.
306
+
307
### ModelScope
308
309
To use models from [ModelScope](https://www.modelscope.cn) instead of Hugging Face Hub, set an environment variable: