# vLLM-Omni: Fully Disaggregated Serving for Any-to-Any Multimodal Models

arXiv: https://arxiv.org/abs/2602.02204 · HTML: https://mlsanity.com/paper/2602.02204/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Feb 2026.

## Built on by (most cited first, 15 of 15)

- [Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators](https://mlsanity.com/paper/2606.06476/) (2026)
- [VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?](https://mlsanity.com/paper/2605.06068/) (2026)
- [dots.tts Technical Report](https://mlsanity.com/paper/2606.07080/) (2026)
- [OmniRefine: Alignment-Aware Cooperative Compression for Efficient Omnimodal Large Language Models](https://mlsanity.com/paper/2605.12056/) (2026)
- [LVSA: Training-Free Sparse Attention for Long Video Diffusion](https://mlsanity.com/paper/2605.31057/) (2026)
- [M*: A Modular, Extensible, Serving System for Multimodal Models](https://mlsanity.com/paper/2606.12688/) (2026)
- [FlashRT: Agent Harness for Guiding Agents to Deploy Real-Time Multimodal Applications](https://mlsanity.com/paper/2607.18171/) (2026)
- [ReDesign: Recovering Editable Design Structures from Images via Agentic Decomposition](https://mlsanity.com/paper/2607.25565/) (2026)
- [VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation](https://mlsanity.com/paper/2610.01499/) (2026)
- [Voxtral TTS](https://mlsanity.com/paper/2603.25551/) (2026)
- [SFL-MTSC: Leveraging Semantic Frame-Level Multi-Task Self-Consistency for Robust Multi-Intent Spoken Language Understanding](https://mlsanity.com/paper/2606.25552/) (2026)
- [Omni-Flow: A Unified Workflow Orchestration and Distributed KV Cache Sharing Framework for Multimodal Inference](https://mlsanity.com/paper/2606.31093/) (2026)
- [LeanGRPO: Eliminating Redundant Recomputation in Diffusion RL](https://mlsanity.com/paper/2609.03528/) (2026)
- [Understanding and Exploiting Diagonal Attention Sparsity in Autoregressive Image Generation](https://mlsanity.com/paper/2609.19702/) (2026)
- [Cache-Aware Conv3D Lowering Across Embedded World-Model Decoders](https://mlsanity.com/paper/2609.31938/) (2026)

## Abstract (arXiv)

Any-to-any multimodal models that jointly handle text, images, video, and audio represent a significant advance in multimodal AI. However, their complex architectures (typically combining multiple autoregressive LLMs, diffusion transformers, and other specialized components) pose substantial challenges for efficient model serving. Existing serving systems are mainly tailored to a single paradigm, such as autoregressive LLMs for text generation or diffusion transformers for visual generation. They lack support for any-to-any pipelines that involve multiple interconnected model components. As a result, developers must manually handle cross-stage interactions, leading to huge performance degradation. We present vLLM-Omni, a fully disaggregated serving system for any-to-any models. vLLM-Omni features a novel stage abstraction that enables users to decompose complex any-to-any architectures into interconnected stages represented as a graph, and a disaggregated stage execution backend that optimizes resource utilization and throughput across stages. Each stage is independently served by an LLM or diffusion engine with per-stage request batching, flexible GPU allocation, and unified inter-stage connectors for data routing. Experimental results demonstrate that vLLM-Omni reduces job completion time (JCT) by up to 91.4% compared to baseline methods. The code is public available at https://github.com/vllm-project/vllm-omni.