tzhouamandClaude Opus 5
67bee59fda
Reading the device profile of the first export showed a fifth of every
sliding layer went nowhere useful: Output 9.9%, its quantise conversion
3.1%, and the window Slice 7.4%. The layer took a 512-entry window in,
appended the new key, sliced the oldest off and handed the whole recomputed
window back out -- 21 of 28 layers paying to move a cache-sized tensor
across the graph boundary once per token. Full layers paid a similar 9.5%
concatenating the new key onto the cache.
Both are the same mistake, and both go away by joining in score space
instead of key space: score the cache and the new token separately, then
concatenate two vectors of length n and 1 rather than an n x 256 key matrix.
Nothing cache-sized is copied, and each layer returns just the new K/V entry
for the runtime to drop into its ring slot. Softmax over keys does not care
what order the slots are in and each cached key already carries its own
rotary phase, so a ring buffer needs no reordering -- only a mask for slots
not yet written. The sliding cache correspondingly holds window-1 entries,
since a 512-token window counts the current token (transformers' own sliding
cache holds 511 for the same reason).
Measured on a real Galaxy S25, fp16, ms per layer:
sliding 1.689 -> 1.483 (-12.2%)
full 1.559 -> 1.567 (unchanged: the concat saved is spent on the
extra score matmul)
Two shapes had to be written as matmuls rather than the obvious elementwise
forms -- the new token's score (broadcast multiply + ReduceSum) and its
value contribution (a two-axis broadcast multiply). Both compile at fp16 and
both make QNN's context-binary converter exit 14 under w4a16; as matmuls
they compile and are the cheaper op on HTP anyway.
Also adds a selectable GELU approximation, and the measurement that closes
the question: erf GELU is 11-13% of every layer, but tanh is *slower* on
this NPU (1.535 vs 1.483 ms) despite holding 65.2 dB. The exact form stays,
which is also the better-fidelity choice.
Parity against the reference implementation is 136.8 dB with identical
top-1/top-5, and a test pins ring and roll to the same output on a shared
history so the device numbers cannot silently measure a different model.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L1ax5wYLA6XbqEL1i8L9gZ
Easy, fast, and cheap omni-modality model serving for everyone
| Documentation | DeepWiki | User Forum | Developer Slack | WeChat | Paper | Slides |
Latest News 🔥
- [2026/08] We released 0.28.0, featuring production-ready MiniMax H3 serving on GPU and NPU, a unified AR/DiT paged KV cache runtime, and enhanced realtime full-duplex serving for the MiniCPM-o series.
- [2026/08] VeRL-Omni
v0.2.0is released: faster diffusion RL powered by vLLM-Omni (request-level/step-wise batching with FA3), rebuilt Qwen3-Omni multimodal training (DPO & GSPO), plus LTX-2.3, Qwen-Image-Edit support and more. See the release notes. - [2026/08] We released 0.26.0 - aligned with the vLLM 0.26 release line, featuring MiniMax H3 joint video/audio generation, an experimental full-duplex realtime runtime for MiniCPM-o 4.5, distributed layerwise diffusion offload, and broader model, hardware, streaming, TTS, and quantization support.
- [2026/07] We released 0.24.0 - aligned with the vLLM 0.24 release line, expanding production-ready coverage across TTS, speech, diffusion, image/video generation, and robot-policy serving, with major Omni stage runtime refactoring, diffusion request-level batching, async output materialization, quantization/cache/memory improvements, and broad CUDA/ROCm/XPU/NPU support.
- [2026/06] Starting with 0.14.0, vLLM-Omni publishes a stable release aligned with every even-numbered upstream vLLM minor version. 0.16.0, 0.18.0, 0.20.0, and 0.22.0 continued this cadence, expanding omni and world-model support with NVIDIA Cosmos3 and DreamZero, adding models such as MiniCPM-o 4.5, MOSS-TTS, and Lance, and advancing TTS, diffusion, distributed execution, quantization, RL integration through VeRL-Omni, and CUDA/ROCm/MUSA/NPU/XPU coverage.
- [2026/03] Check out our first public project deepdive at the vLLM Hong Kong Meetup!
- [2025/11] vLLM community officially released vllm-project/vllm-omni in order to support omni-modality models serving.
About
vLLM was originally designed to support large language models for text-based autoregressive generation tasks. vLLM-Omni is a framework that extends its support for omni-modality model inference and serving:
- Omni-modality: Text, image, audio, video, and action data processing
- Non-autoregressive Architectures: extend the AR support of vLLM to Diffusion Transformers (DiT) and other parallel generation models
- Heterogeneous outputs: from traditional text generation to multimodal and action outputs
vLLM-Omni is fast with:
- State-of-the-art AR support by leveraging efficient KV cache management from vLLM
- Pipelined stage execution overlapping for high throughput performance
- Fully disaggregation based on OmniConnector and dynamic resource allocation across stages
vLLM-Omni is flexible and easy to use with:
- Heterogeneous pipeline abstraction to manage complex model workflows
- Seamless integration with popular Hugging Face models
- Tensor, pipeline, data and expert parallelism support for distributed inference
- Streaming outputs
- OpenAI-compatible API server
- Full-duplex realtime serving with streaming audio input and output
vLLM-Omni seamlessly supports most popular open-source models on HuggingFace, including:
- Omni-modality models (e.g. Qwen3-Omni, MiniCPM-o 4.5, Cosmos3, HunyuanImage, BAGEL)
- TTS models (e.g. Qwen3-TTS, IndexTTS 2.5, dots.tts, CosyVoice3)
- Diffusion models — image, video, and audio generation (e.g. MiniMax H3, LTX-2.5, SANA-Video, Wan2.2)
- Robot-policy and action models (e.g. π0, GR00T-N1.7, DreamZero-DROID, InternVLA-A1)
Getting Started
Visit our documentation to learn more.
- Installation
- Quickstart
- Nightly CUDA Docker images built from the latest
main - List of Supported Models
- Deployment Recipes for vLLM-Omni model serving
Contributing
We welcome and value any contributions and collaborations. Please check out Contributing to vLLM-Omni for how to get involved.
Citation
If you use vLLM-Omni for your research, please cite our paper:
@article{yin2026vllmomni,
title={vLLM-Omni: Fully Disaggregated Serving for Any-to-Any Multimodal Models},
author={Peiqi Yin, Jiangyun Zhu, Han Gao, Chenguang Zheng, Yongxiang Huang, Taichang Zhou, Ruirui Yang, Weizhi Liu, Weiqing Chen, Canlin Guo, Didan Deng, Zifeng Mo, Cong Wang, James Cheng, Roger Wang, Hongsheng Liu},
journal={arXiv preprint arXiv:2602.02204},
year={2026}
}
Join the Community
Feel free to ask questions, provide feedbacks and discuss with fellow users of vLLM-Omni in #sig-omni slack channel at slack.vllm.ai or vLLM user forum at discuss.vllm.ai.
Star History
License
Apache License 2.0, as found in the LICENSE file.