Is this page helpful?

Release Notes#

Release 2.1.2#

This release contains the following update.

SGLang Model-Free NIM#

This is an updated release of the SGLang Model-Free NIM, which lets you run any model supported by SGLang without a model-specific container image. For more information, refer to Model-Free NIM.

Note the following SGLang Model-Free NIM limitations:

  • The environment variable NIM_MAX_MODEL_LEN is not supported by this container.

  • Video input requires mounting FFmpeg8 into the container. Add the following volume mount to your docker run command:

    -v /path/to/ffmpeg8/lib:/opt/ffmpeg8:ro \

  • thinking=false is mapped to reasoning_effort=low. Reasoning is hidden from the response but still consumes output tokens. If you configure a stop word, it could match the stop words in the hidden reasoning field.

For information about past updates and older versions, refer to the previous release notes.