Is MOSS-VL-Realtime-NF4 supported by SGLang?

[WindLWQ]

Hello,

I would like to ask whether MOSS-VL-Realtime-NF4 is currently supported by SGLang.

We are planning to deploy MOSS-VL-Realtime-NF4 for inference and would like to use SGLang as the inference backend.

Could you please clarify:

Is MOSS-VL-Realtime-NF4 compatible with SGLang?
If so, is there any recommended configuration, branch, or specific SGLang version that we should use?
If it is not currently supported, are there any plans to add SGLang support in the future?

Thank you for your help!

[SSSSuperC]

Hello,

Thank you for your interest in MOSS-VL-Realtime-NF4.

Unfortunately, SGLang inference is currently not supported for any MOSS-VL Realtime variant — this includes the
NF4-quantized version you mentioned.

At the moment, we recommend using the official inference code provided in the repository for deployment.

We do plan to work on SGLang adaptation in the future. Once support is ready, we will announce it in the repository.

Thank you for your understanding, and feel free to reach out if you have any other questions.