vLLM adds serving support for a hybrid Mamba-Transformer MoE vision-language model

vLLM says it now serves a hybrid Mamba-Transformer mixture-of-experts model paired with a C-RADIOv4 vision encoder.

The serving stack includes multi-token prediction (MTP) for speculative decoding, continuous batching and prefix caching. vLLM says the setup provides visual understanding and long-context reasoning through an OpenAI-compatible endpoint, for document, research and video workflows.

The announcement comes from vLLM's project account on X, dated October 15, 2026. The report does not include benchmark results or the model's name.