v0.18.0

## Highlights

This release features 324 commits from 83 contributors, including 38 new contributors.

vLLM-Omni v0.18.0 is a major rebase and systems release that aligns the project with upstream vLLM v0.18.0, strengthens the core runtime through a large entrypoint refactor and scheduler/runtime cleanups, expands unified quantization and diffusion execution, broadens multimodal model coverage, and improves production readiness across audio, omni, image, video, RL, and multi-platform deployments. 

### Key Improvements

* **Rebased to upstream vLLM v0.18.0**, with follow-up updates to docs and dockerfiles, plus cleanup of patches that were no longer needed after the rebase. **(#2037, #2038, #2062, #2271)** 
* **Refactored the serving entrypoint architecture**, making the stack cleaner and easier to extend, while also laying groundwork for PD disaggregation, multimodal output decoupling, coordinator-based orchestration, and pipeline config cleanup. **(#1908, #1863, #1816, #1465, #1115)** 
* **Strengthened audio, speech, and omni production serving**, especially for Qwen3-TTS, Qwen3-Omni, MiMo-Audio, Fish Speech S2 Pro, and Voxtral TTS, with lower latency, better concurrency, more robust streaming, and improved online serving stability. **(#1583, #1617, #1797, #1913, #1985, #1852, #1656, #1963, #2009, #2019, #2239, #1688, #1752, #1964, #2225, #1859, #2145, #2151, #2156, #2158)** 
* **Delivered substantial diffusion optimization**, with scheduler/executor refactoring, faster startup, better cache-dit / TeaCache integration, broader TP/SP/HSDP support, and multiple correctness fixes for online and offline serving. **(#1625, #1504, #1715, #1834, #1848, #1234, #2163, #1979, #2101, #2176)** 
* **Expanded model support across omni, speech, image, and video**, including Helios, Helios-Mid / Distilled, MammothModa2, Fun CosyVoice3-0.5B-2512, FLUX.2-dev, FLUX.1-Kontext-dev, Hunyuan Image3 AR, Fish Speech S2 Pro, Voxtral TTS, DreamID-Omni, LTX-2, and HunyuanVideo-1.5. **(#1604, #1648, #336, #498, #1629, #561, #759, #1798, #1803, #1855, #841, #1516)** 
* **Introduced a unified quantization framework** and expanded quantization support across diffusion and image workloads, including INT8, FP8, and GGUF-related enablement. **(#1764, #1470, #1640, #1755, #1473, #2180)** 
* **Improved RL and custom pipeline readiness, verl collaboration & Qwen-Image E2E RL**, Expanded RL-oriented serving in close collaboration with verl, helping enable Qwen-Image end-to-end RL / Flow-GRPO training with collective RPC support. Including collective RPC support at the entrypoint, custom input/output support, async batching for Qwen-Image, and dedicated E2E coverage for custom RL pipelines. **(#1646, #1593, #2005, #2217)** 

### Core Architecture & Runtime

* Reworked the core serving architecture through the **vLLM-Omni Entrypoint Refactoring**, while also adding PD disaggregation scaffolding, coordinator support, multimodal output decoupling foundations, and cleaner model/pipeline configuration handling. **(#1908, #1863, #1465, #1816, #1115, #1958, #2105)** 
* Continued cleanup of runtime internals with stage/step pipeline refactors, dead-code cleanup, and improvements to async engine robustness and scheduler state handling. **(#1368, #1579, #2153, #2028, #1893)** 

### Model Support

* **Omni / speech / audio models:** added or expanded support for **MammothModa2**, **Fun CosyVoice3-0.5B-2512**, **Fish Speech S2 Pro**, and **Voxtral TTS**. **(#336, #498, #1798, #1803)** 
* **Image / diffusion models:** added or expanded support for **Hunyuan Image-3.0**, **FLUX.2-dev**, **FLUX.1-Kontext-dev**, and continued improvements for **Qwen-Image**, **Qwen-Image-Edit**, **Qwen-Image-Layered**, **LongCat-Image**, **GLM-Image**, **Bagel**, and **OmniGen2**. **(#759, #1629, #561, #1682, #2085, #1970, #2035, #1918, #1578, #1669, #1903, #1711, #1934)** 
* **Video models:** added or expanded support for **Helios**, **Helios-Mid / Distilled**, **DreamID-Omni**, **LTX-2**, **HunyuanVideo-1.5**, and updated supported video-generation coverage for **Wan2.1-T2V**. **(#1604, #1648, #1855, #841, #1516, #1920)** 

### Audio, Speech & Omni Production Optimization

* **Qwen3-TTS** received major optimization work, including lower TTFA, better high-concurrency throughput, improved Code Predictor / Code2Wav execution, websocket streaming audio output, async scheduling by default, voice upload support, optional `ref_text`, and long `ref_audio` handling fixes. **(#1583, #1617, #1797, #1913, #1985, #1852, #1719, #1853, #1201, #1879, #2046, #2104)** 
* **Qwen3-Omni** gained lower inter-packet latency, speaker-switching support, decode-alignment fixes, and multiple correctness fixes for answer quality and online serving stability. **(#1656, #1963, #2009, #2019, #2239)** 
* **MiMo-Audio** improved compatibility and production robustness with TP fixes, broader attention backend support, configurable chunk sizing, and documentation to prevent noise-only outputs under unsupported attention setups. **(#1688, #1752, #1964, #2225, #2205)** 
* **Fish Speech S2 Pro** and **Voxtral TTS** were productionized further with online serving, voice cloning, better TTFP / inference performance, multilingual demo support, lighter flow matching, and voice-embedding fixes. **(#1798, #1859, #2145, #1803, #2045, #2056, #2067, #2151, #2156, #2158, #2023)** 
* Added or improved speech-serving interfaces, including **speech batch entrypoint**, **speaker embedding support for speech and voices APIs**, proper HTTP status handling, and streaming `wav` response support. **(#1701, #1227, #1687, #1819)** 

### Diffusion, Image & Video Generation

* **Runtime refactor & benchmarking:** Refactored the diffusion runtime with cleaner scheduler/executor boundaries, better request-state flow, unified profiling, and stronger benchmarking infrastructure. **(#1625, #2099, #1757, #1917, #1995)**
* **Performance & startup gains:** Improved diffusion performance through multi-threaded weight loading for Wan2.2, reduced IPC overhead for single-stage serving, cache-dit upgrades, TeaCache support, and nightly performance improvements for Qwen-Image. **(#1504, #1715, #1834, #1234, #1314, #1805, #2111)**
* **Distributed scaling:** Expanded distributed diffusion execution with broader TP/SP/HSDP support across Flux, GLM-Image, Hunyuan, and Bagel. **(#1250, #1900, #1918, #2163, #1903)**
* **Serving UX & API ergonomics:** Improved serving usability with a progress bar for diffusion models, richer image-edit parameters such as layers and resolution, and extra request-body support for video APIs. **(#1652, #2053, #1955)**
* **Correctness & stability fixes:** Fixed a wide range of diffusion correctness issues, including config misalignment between offline and online inference, TP/no-seed broken-image issues, GLM-Image stage/device bugs, and TeaCache incompatibilities. **(#1979, #2176, #2137, #2101, #1894, #2025)**


### Quantization & Memory Efficiency

* Added the **Unified Quantization Framework** as a core infrastructure upgrade for more consistent quantized execution across model families. **(#1764)** 
* Expanded quantization support for diffusion/image workloads, including **INT8 for DiT (Z-Image and Qwen-Image)**, **FP8 for Flux transformers**, and **GGUF adapter support for Qwen-Image**. **(#1470, #1640, #1755)** 
* Improved compatibility between quantization and runtime features such as **CPU offload**, **tensor parallelism**, and **Flux-family execution**. **(#1473, #1723, #1978, #2180)** 

### RL, Serving & Integrations

* **verl collaboration & Qwen-Image E2E RL:** Expanded RL-oriented serving in close collaboration with verl, helping enable Qwen-Image end-to-end RL / Flow-GRPO training with collective RPC support, custom input/output, async batching for Qwen-Image, and dedicated E2E CI coverage for custom RL pipelines. **(#1646, #1593, #2005, #2217)**
* **Rollout scaling for visual RL:** Added rollout building blocks referenced by verl’s Qwen-Image integration plan, including async batching for Qwen-Image plus tensor-parallel and data-parallel support for diffusion serving. **(#1593, #1713, #1706)**
* **Deployment & ecosystem integrations:** Improved deployment and ecosystem integration with a Helm chart for Kubernetes, ComfyUI video & LoRA support, and a rewritten async video API lifecycle. **(#1337, #1596, #1665)**

### Platforms, Distributed Execution & Hardware Coverage

* Continued improving portability across **CUDA, ROCm, NPU, and XPU/Intel GPU** environments, including rebase follow-ups, ROCm CI setup, Intel CI dispatch, Intel GPU docs, and NPU docker/docs refreshes. **(#2017, #1984, #1721, #2154, #2271, #2091)** 
* Expanded distributed execution coverage with **T5 tensor parallelism**, more model-level TP/SP/HSDP support, and better handling of visible GPUs and stage-device initialization. **(#1881, #1250, #1900, #1918, #2163, #2025)** 

### CI, Benchmarks & Documentation

* Strengthened release engineering and CI with a **release pipeline**, richer nightly benchmark/report generation, L3/L4/L5 test layering, expanded model E2E coverage, and stronger diffusion test coverage. **(#1726, #1831, #1995, #1514, #1799, #2086, #1869, #2085, #2087, #2132, #2129, #2023)** 
* Improved benchmarking with **Qwen3-TTS benchmark scripts**, nightly Qwen3-TTS and Qwen-Image performance tracking, diffusion timing, random benchmark datasets, and T2I/I2I accuracy benchmark integration. **(#1573, #1700, #1805, #2111, #1757, #1657, #1917)** 
* Refreshed project docs across installation, omni/TTS docs, diffusion serving parameters, UAA documentation, developer guides, and governance. **(#1762, #1693, #2051, #2130, #2148, #1889)** 

### Note
  * GLM-Image requires manually upgrading the `transformers` version to >= 5.0.

## What's Changed
* 0.16.0 release by @ywang96 in https://github.com/vllm-project/vllm-omni/pull/1576
* [Refactor]: Phase1 for rebasing_additional_info by @divyanshsinghvi in https://github.com/vllm-project/vllm-omni/pull/1394
* [Feature]: Support cfg kv-cache transfer in multi-stage by @princepride in https://github.com/vllm-project/vllm-omni/pull/1422
* [BugFix] Fix load_weights error when loading HunyuanImage3.0 by @Semmer2 in https://github.com/vllm-project/vllm-omni/pull/1598
* [Bugfix] fix kernel error for qwen3-omni by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/1602
* [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/1570
* [Bugfix] Import InputPreprocessor into Renderer by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1566
* [Feature][Wan2.2] Speed up diffusion model startup by multi-thread weight loading by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1504
* [Bugfix][Model] Fix LongCat Image Config Handling / Layer Creation by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1485
* [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context by @ZhanqiuHu in https://github.com/vllm-project/vllm-omni/pull/1619
* [Debug] Enable curl retry aligned with openai by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/1539
* [Doc] Fix links in the configuration doc by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/1615
* [CI] Add scripts for bechmark collection and email distribution. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1307
* [FEATURE] Tile/Patch parallelism refactor for easily support other models by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/1366
* [Bugfix] Fix filepath resolution for model with subdir and GLM-Image generation by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/1609
* Make chunk_size and left_context_size configurable via YAML for async chunking by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/1423
* [Bugfix] Fix transformers 5.x compat issues in online TTS serving by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1536
* [Refactor] lora: reuse load_weights packed mapping by @dongbo910220 in https://github.com/vllm-project/vllm-omni/pull/991
* [Model]: Support Helios  from ByteDance by @princepride in https://github.com/vllm-project/vllm-omni/pull/1604
* [chore] add _repeated_blocks for regional compilation support by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1642
* [Bugfix] Add TTS request validation to prevent engine crashes by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1641
* [CI] Fix ASCII codes. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1647
* [Misc] update wechat by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1649
* docs: Announce vllm-omni-skills community project by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/1651
* [Model] Add Hunyuan Image3 AR Support by @usberkeley in https://github.com/vllm-project/vllm-omni/pull/759
* [Test][Qwen3-Omni]Modify Qwen3-Omni benchmark test cases by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/1628
* [Bugfix] Fix Dtype Parsing by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1391
* [XPU] fix UMD version in docker file by @yma11 in https://github.com/vllm-project/vllm-omni/pull/1545
* add support for MammothModa2 model by @HonestDeng in https://github.com/vllm-project/vllm-omni/pull/336
* [Model] Fun cosy voice3-0.5-b-2512 by @divyanshsinghvi in https://github.com/vllm-project/vllm-omni/pull/498
* [Bugfix] Enable torch.compile for low noise model (transformer_2) by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1541
* [NPU] [Features] [Bugfix] Support mindiesd adaln by @jiangmengyu18 in https://github.com/vllm-project/vllm-omni/pull/1537
* [FP8 Quantization] Add FP8 quantization support for Flux transformer by @zzhuoxin1508 in https://github.com/vllm-project/vllm-omni/pull/1640
* Replace hard-coded cuda generator with current_omni_platform.device_type by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/1677
* [BugFix] Fix LongCat Sequence Parallelism / Small Cleanup by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1631
* [Misc] remove logits_processor_pattern this field, because vllm have … by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1675
* [CI] Remove high concurrency tests before issue #1374 fixed. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1683
* [Optimize][Qwen3-Omni] Reduce inter-packet latency in async chunk  by @ZeldaHuang in https://github.com/vllm-project/vllm-omni/pull/1656
* [Feat][Qwen3TTS] reduce TTFA with flexible initial phase by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1583
* [Model] support LTX-2 text-to-video image-to-video by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/841
* [BugFix] Return proper HTTP status for ErrorResponse in create_speech by @Lidang-Jiang in https://github.com/vllm-project/vllm-omni/pull/1687
* [Doc] Add the test guide document. [skip ci] by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1376
* [UX] Add progress bar for diffusion models by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1652
* [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder by @ZhanqiuHu in https://github.com/vllm-project/vllm-omni/pull/1664
* [Feature] Support flexible task_type configuration for Qwen3-TTS models by @JackLeeHal in https://github.com/vllm-project/vllm-omni/pull/1197
* [Cleanup] Move cosyvoice3 tests to model subdirectory by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1666
* [Feature][Bagel] Add CFG parallel mode by @nussejzz in https://github.com/vllm-project/vllm-omni/pull/1578
* perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor by @dubin555 in https://github.com/vllm-project/vllm-omni/pull/1614
* [Refactor][Perf] Qwen3-TTS: re-prefill Code Predictor with torch.compile + enable Code2Wav decoder CUDA Graph by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/1617
* [MiMo-Audio] Bugfix tp lg than 1 by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/1688
* Add non-async chunk support for Qwen3-TTS by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1678
* [1/N][Refactor] Clean up dead code in output processor by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1579
* [feature]: support flux2.klein cache_dit by @nuclearwu in https://github.com/vllm-project/vllm-omni/pull/1209
* [skip CI][Docs] Add TTS model developer guide by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1693
* [Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline by @erfgss in https://github.com/vllm-project/vllm-omni/pull/668
* [Feature]:  Add vae-patch-parallel CLI argument in online serving by @wtomin in https://github.com/vllm-project/vllm-omni/pull/1716
* Revert "[Profile] Adding metrics for Diffusion/DiT Single diffusion Pipeline (#668)" by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1724
* [CI] Add release-pipeline.yaml. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1726
* [NPU] Support Helios-Mid / Distilled by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1648
* [skip ci] Update slides link by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/1730
* [Bugfix] (qwen3_tts): enable batched offline inference by fixing tens… by @RomanKoshkin in https://github.com/vllm-project/vllm-omni/pull/1417
* [Bugfix] Use upstream MediaConnector for ref_audio resolution by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1661
* [RL] Support collective rpc api to entrypoint && Support custom input output  by @knlnguyen1802 in https://github.com/vllm-project/vllm-omni/pull/1646
* Pre-download Qwen3-TTS model in CI to avoid intermittent download timeouts by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1727
* [1/N] fix CP for Helios by @SHYuanBest in https://github.com/vllm-project/vllm-omni/pull/1729
* feat(tts): add voice upload API for Qwen3-TTS by @zhaotyer in https://github.com/vllm-project/vllm-omni/pull/1201
* [Bagel] Eliminate broadcast in CFG parallel denoising loop by @nussejzz in https://github.com/vllm-project/vllm-omni/pull/1695
* [Feat]: Offline inference supports async_chunk by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/1415
* [Bugfix] Allow to enable HSDP alone by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1567
* Disable mm processor cache in CI stage configs by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1739
* Dev/rebase v0170 by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/1639
* [Perf] Reduce IPC overhead for single-stage diffusion serving for Wan2.2 by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1715
* [Test] Solving the Issue of Whisper Model's GPU Memory Not Being Successfully Cleared and the Occasional Accuracy Problem of the Qwen3-omni Model Test by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1744
* [Bagel]: Support multistage img2img by @princepride in https://github.com/vllm-project/vllm-omni/pull/1669
* [BugFix] Enable CPU offloading and Cache-DiT together on Diffusion Model by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/1723
* [Doc] CLI Args Naming Style Correction by @wtomin in https://github.com/vllm-project/vllm-omni/pull/1750
* [Feature] Add Helm Chart to deploy vLLM-Omni on Kubernetes by @oglok in https://github.com/vllm-project/vllm-omni/pull/1337
* [Fix][Qwen3-TTS] Preserve ref_code decoder context for Base ICL by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/1731
* Add online serving to Stable Audio Diffusion and introduce `v1/audio/generate` endpoint by @ekagra-ranjan in https://github.com/vllm-project/vllm-omni/pull/1255
* [Enhancement][pytest] Check for process running during start server by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/1559
* [CI]: Add core_model and cpu markers for L1 use case. by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/1709
* [Doc][skip-ci] Update installation instructions by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/1762
* Revert "Add online serving to Stable Audio Diffusion and introduce `v1/audio/generate` endpoint" by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/1789
* [BUGFIX] Add compatibility for mimo-audio with vLLM 0.17.0 by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/1752
* [feat][Qwen3TTS] Simple dynamic TTFA based on Code2Wav load by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1714
* [Refactor][Perf] Qwen3-omni: code predictor with re-prefill + SDPA and eliminate decode hot-path CPU round-trips by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/1758
* [Feat][Qwen3-tts]: Add Gradio demo for online serving by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1231
* [Docs] update async chunk performance diagram by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/1741
* [Feat] Enable expert parallel for diffusion MoE layers by @Semmer2 in https://github.com/vllm-project/vllm-omni/pull/1323
* [Bugfix]: SP attention not enabling when _sp_plan hooks are not applied by @wtomin in https://github.com/vllm-project/vllm-omni/pull/1704
* [skip ci] [Docs] Update WeChat QR code for community support by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1802
* update GpuMemoryMonitor to DeviceMemoryMonitor for all HW by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/1526
* Add coordinator module and corresponding unit test by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/1465
* [Model]: add FLUX.2-dev model by @nuclearwu in https://github.com/vllm-project/vllm-omni/pull/1629
* [skip ci][Docs] doc fix for example snippets by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1811
* [Test] L4 complete diffusion feature test for Qwen-Image-Edit models by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/1682
* [Frontend] ComfyUI video & LoRA support by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/1596
* [Bugfix] Adjust Z-Image Tensor Parallelism Diff Threshold by @wtomin in https://github.com/vllm-project/vllm-omni/pull/1808
* [Bugfix] Expose base_model_paths property in _DiffusionServingModels by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1771
* [Bugfix] Report supported tasks for omni models to skip unnecessary chat init by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1645
* [Test] Add Qwen3-TTS nightly performance benchmark by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1700
* Add Qwen3-TTS benchmark scripts by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1573
* [Test] Skip the qwen3-omni relevant validation for a known issue 1367. by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1812
* Fix duplicate get_supported_tasks definition in async_omni.py by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1825
* [Enhancement] Patch OmniStage.try_collect() with _proc alive checks by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/1560
* [Doc][skip ci] Update readme with Video link for vLLM HK First Meetup by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1833
* [Feat][Qwen3-TTS] Support streaming audio output for websocket  by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/1719
* [Test] Nightly Buildkite Pytest Test Case Statistics And Send HTML Report By Email by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1674
* [Enhancement] Patch OmniStage.try_collect() with ray alive checks by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/1561
* [Feat][Diffusion]: Implement Component-Level VRAM Quota and Resource Domain Isolation by @Flink-ddd in https://github.com/vllm-project/vllm-omni/pull/1582
* [Feature]: Enable directly use OmniLLM init AR model by @princepride in https://github.com/vllm-project/vllm-omni/pull/1821
* [Enhancement] Upgrade cache-dit from 1.2.0 to 1.3.0 by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1834
* [Bugfix] Modify _resolve_pytest_target to support glob patterns and return multiple paths by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1843
* [Feat] add wav response_format when stream is true in /v1/audio/speec… by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1819
* [BugFix]: Revert #1582 by @princepride in https://github.com/vllm-project/vllm-omni/pull/1842
* [Feature]: support Flux.2-dev cache_dit by @nuclearwu in https://github.com/vllm-project/vllm-omni/pull/1814
* [skip ci] update readme slides link by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/1850
* [Model] Extend NPU support for HunyuanImage3 Diffusion Model by @ElleElleWu in https://github.com/vllm-project/vllm-omni/pull/1689
* [Config Refactor][1/2] Model Pipeline Configuration System by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1115
* [Test] Reduce SP & Offloading test cases for L2 by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/1839
* [bugfix] Add Interleaved 2D Rotary Embedding for HunyuanImage3 by @usberkeley in https://github.com/vllm-project/vllm-omni/pull/1784
* [Bugfix] Fix Helios text_encoder embed_tokens all-zeros due to untied weights by @dubin555 in https://github.com/vllm-project/vllm-omni/pull/1728
* Enable async_scheduling by default for Qwen3-TTS by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1853
* [CI failure] Comment out test_zimage_vae_patch_parallel_tp2 by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/1856
* Add Fish Speech S2 Pro support with online serving and voice cloning by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1798
* [skip CI][Docs] add connector design document by @natureofnature in https://github.com/vllm-project/vllm-omni/pull/1737
* [BugFix] Readme and example runner file for cosyvoice3 missed in refactoring by @divyanshsinghvi in https://github.com/vllm-project/vllm-omni/pull/1685
* [Refactor] Use SP Plan for LongCat Sequence Parallelism by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1772
* [CI failed] Disable test for zimage tensor parallelism by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/1870
* [Bugfix] Fix SD3.5-medium attn2 uninitialized weights by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1659
* [Bugfix] fix layer-wise offload incompatible with cache-dit by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1786
* [CI failed] Disable Diffusion Tensor Parallelism Test by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/1876
* [BugFix]: Fix bagel online inference bug by @princepride in https://github.com/vllm-project/vllm-omni/pull/1804
* [Frontend] Rewrite video API for async job lifecycle by @ieaves in https://github.com/vllm-project/vllm-omni/pull/1665
* [Diffusion] [Model] Dreamid-Omni from bytedance by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/1855
* [Bugfix] Restore voice upload API and profiler endpoints reverted by #1719 by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1879
* [BugFix] Fix Max Rank Handling in LoRA by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1397
* Buildkite hardware ci xpu test by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/1340
* [CI] add multimodal processing correctness tests for Omni models by @zzhuoxin1508 in https://github.com/vllm-project/vllm-omni/pull/1445
* fix: propagate parallel_config through create_default_diffusion by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1878
* [CI pipeline] Re-enable Diffusion Tensor Parallelism Test in pipeline by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/1892
* [skip CI][Docs][Benchmark]: clarify vbench parameter behavior and add t2v example by @asukaqaq-s in https://github.com/vllm-project/vllm-omni/pull/1497
* [Bugfix] Fix cpu offload and quantization compatibility by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1473
* [Feat] support SP for FLUX.2-klein by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1250
* [CI]: Add/Fix bagel e2e online/offline test by @princepride in https://github.com/vllm-project/vllm-omni/pull/1895
* [Feat] support HSDP for Flux family by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1900
* Add `Governance` section by @ywang96 in https://github.com/vllm-project/vllm-omni/pull/1889
* Update latest news section in README.md by @ywang96 in https://github.com/vllm-project/vllm-omni/pull/1909
* [Feature] Split #1303 Part 1: PD disaggregation scaffolding by @ahengljh in https://github.com/vllm-project/vllm-omni/pull/1863
* [NPU] Upgrade to v0.17.0 by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1890
* [Misc] removed qwen3_tts.py as it is out-dated by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1926
* [Bug][Qwen3TTS][Streaming] remove dynamic initial chunk and only compute on initial request by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1930
* Fix Base voice clone streaming quality and stop-token crash by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1945
* [Docs] Update WeChat QR code for community support by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1974
* [skip ci][Docs] Update WeChat QR code (fix filename case) by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1976
* [Entrypoint][Refactor] vLLM-Omni Entrypoint Refactoring by @fake0fan in https://github.com/vllm-project/vllm-omni/pull/1908
* [Bugfix] Set PREEMPTED status when moving requests from running to waiting queue by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/1893
* [Feature] Add cache-dit support for HunyuanImage3 by @Fishermanykx in https://github.com/vllm-project/vllm-omni/pull/1848
* [Feature]: Remove some useless `hf_overrides` in yaml by @princepride in https://github.com/vllm-project/vllm-omni/pull/1898
* [CI] Nightly Benchmark - Add an HTML generator, Update the EXCEL generator. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1831
* [Bug]: fix CUDA OOM during diffusion post-processing by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1670
* [Optim][Qwen3TTS] big boost model throughput+latency high concurrency by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1852
* [CI] [ROCm] Bugfix device environment issue by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/1984
* [CI]init intel ci dispatch in buildkite folder by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/1721
* Fix OmniGen2 transformer config loading for HF models by @Joshna-Medisetty in https://github.com/vllm-project/vllm-omni/pull/1934
* [Test] L4 complete diffusion feature test for Bagel models by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/1938
* [Performance] diffusion timing by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/1757
* [Perf] [Qwen3-TTS] Keep audio_codes and last_talker_hidden on GPU to eliminate per-step sync stalls by @DomBrown in https://github.com/vllm-project/vllm-omni/pull/1985
* [CI] Split BAGEL tests into dummy/real weight tiers (L2/L3) by @princepride in https://github.com/vllm-project/vllm-omni/pull/1998
* [Bugfix] Fix config misalignment between offline and online diffusion inference (Wan2.2, Qwen-Image series) by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1979
* Add HF token to H100 jobs by @khluu in https://github.com/vllm-project/vllm-omni/pull/2008
* [Bugfix] Fix Ovis Image crash when guidance_scale is set without negative_prompt by @Dnoob in https://github.com/vllm-project/vllm-omni/pull/1956
* [Bugfix] fix helios video generate use cpu device by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1915
* [XPU] update bagel modeling to remove cuda hardcode, add xpu stage_config by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/1931
* [Fix] Fix slow hasattr in CUDAGraphWrapper.__getattr__ by @ZeldaHuang in https://github.com/vllm-project/vllm-omni/pull/1982
* [Bugfix] revert PR#1758 which introduced the accuracy problem of qwen3-omni by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/2009
* [Bugfix]Fix bug of online server can not return mutli images by @Hu1Lcode in https://github.com/vllm-project/vllm-omni/pull/2007
* [CI] [ROCm] Setup `test-ready.yml` and `test-merge.yml` by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/2017
* Int8 Quantization Support for DiT (Z-Image & Qwen-Image) by @yjb767868009 in https://github.com/vllm-project/vllm-omni/pull/1470
* [Model] Add Voxtral TTS model by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/1803
* [Feat] Support T5 Tensor Parallelism by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/1881
* [Feat][Qwen3TTS][Code2wav] triton SnakeBeta and Cuda Graph by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1797
* [Optim][Qwen3TTS][CodePredictor] support torch.compile with reduce-overhead and dynamic False by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/1913
* [CI] Change Bagel online test environment variable `VLLM_TEST_CLEAN_GPU_MEMORY` to `0` by @princepride in https://github.com/vllm-project/vllm-omni/pull/2032
* [BugFix][Doc]Update voxtral_tts end2end.py & README.md by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2026
* [Docs] Add Wan2.1-T2V as supported video generation models by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1920
* [Bugfix] Remove duplicated config keyword max batch size by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/1851
* [Test] Implement mock HTTP request handling in benchmark CLI tests by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/2014
* [CI] Fix test. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/2031
* reafator pipeline stage/step pipeline by @asukaqaq-s in https://github.com/vllm-project/vllm-omni/pull/1368
* [Fixbug][Perf] Qwen3-omni: code predictor with re-prefill + SDPA and eliminate decode hot-path CPU round-trips by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/2012
* [Benchmark] [Diffusion] [Enhancement] Random dataset by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/1657
* [Bugfix] Z-Image CFG threshold should be > 0 instead of > 1 by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1634
* [Voxtral TTS] Remove redundant yaml by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2056
* [Bugfix]: fixed ServerDisconnectedError in benchmark test (reapply #1683, fixes #1374) by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/1841
* [Perf] Improve Fish Speech S2 Pro inference performance by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/1859
* [Voxtral] Improve example by @patrickvonplaten in https://github.com/vllm-project/vllm-omni/pull/2045
* [CI] Uncomment condition for nightly build in YAML by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/2057
* [bugfix] /chat/completion doesn't read extra_body for diffusion model by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/2042
* [BugFix][Qwen3TTS] CodePredictor CudaGraph Pool by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/2059
* [Rebase] Rebase to vllm v0.18.0 by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/2037
* [Doc] Update docs and dockerfiles for rebase of vllm v0.18.0 by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/2038
* [Model] Add HunyuanVideo-1.5 T2V and I2V support by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1516
* [Bugfix] Fix Fish Speech and CosyVoice3 online serving - missing is_comprehension and broken model detection by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2058
* Remove mm_prefix_lm patch because vllm==0.18.0 already support by @princepride in https://github.com/vllm-project/vllm-omni/pull/2062
* [Bugfix] Fix HunyuanVideo-1.5 CI failures by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/2066
* [Voxtral] Fix Voxtral TTS end2end.py by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2067
* [FP8] enable hunyuan-image-3 diffusion model with fp8 online quant by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/1935
* [CI] Add Flux2 Klein Tests by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/2027
* [Bugfix] Restore chunk-waiting requests on OmniNewRequestData rewrap failure by @dubin555 in https://github.com/vllm-project/vllm-omni/pull/1691
* [Fix] Fix non-unique request IDs in /v1/images/edits endpoint by @zJuuu in https://github.com/vllm-project/vllm-omni/pull/2050
* [Bugfix] Fix cache-dit for single-transformer Wan2.2 models(eg. Wan2.2-TI2V-5B) by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1392
* [Core] Simplify OmniModelConfig Initialization by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1768
* Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in https://github.com/vllm-project/vllm-omni/pull/2070
* [BugFix]: Fix OmniGen2 Model Loading by @legitnull in https://github.com/vllm-project/vllm-omni/pull/1711
* [Feat] support TeaCache for Flux2 klein by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1234
* [Feature] add Tensor Parallelism to Omnigen2 by @zzhuoxin1508 in https://github.com/vllm-project/vllm-omni/pull/2065
* [Bugfix] fix gguf TypeError: GGUFConfig.get_name() missing 1 required positional argument: 'self', test: add diffusion gguf unit coverage by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1865
* [Docs][CI] doc update & L4 example test for text-to-image page by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/1910
* [Bugfix] Fix NPU Hunyuan fused MoE forward context after rebase to 0.18.0 by @Fishermanykx in https://github.com/vllm-project/vllm-omni/pull/2091
* [Feature][RL] Support batching for QwenImage in async mode by @knlnguyen1802 in https://github.com/vllm-project/vllm-omni/pull/1593
* [Test] L5 Long-Term Stability Test and GPU Memory Monitoring Main L5 last by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/1799
* [CI] Update Diffusion Model Test Configuration for Nightly Builds by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/2086
* [Refactor] Refactor Diffusion Scheduler/Executor Boundaries and Request State Flow by @yJader in https://github.com/vllm-project/vllm-omni/pull/1625
* [Quantization] feat: add qwen-image gguf adapter by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1755
* [Bagel]: Support SP by @princepride in https://github.com/vllm-project/vllm-omni/pull/1903
* [Feat] Phase 1 foundation types for multimodal output decoupling by @meghaagr13 in https://github.com/vllm-project/vllm-omni/pull/1816
* [Unit Test] add unit tests for AsyncOmni and Omni by @yinpeiqi in https://github.com/vllm-project/vllm-omni/pull/2034
* [Perf] Qwen-Image Performance Nightly CI test by @wtomin in https://github.com/vllm-project/vllm-omni/pull/1805
* [Bugfix] Fix Qwen-Image SP and TeaCache incompatibility by @wtomin in https://github.com/vllm-project/vllm-omni/pull/2101
* [Bugfix][Chunk Transfer Adapter] deque mutated fix  by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/2102
* [model] support FLUX.1-Kontext-dev by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/561
* [Doc] Sync and fix. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/2110
* [Feat] support TP for GLM-Image by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1918
* [Feat][Qwen3-TTS] Better Qwen3-TTS online serving demo by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1857
* Add tool to configure gpu_memory_utilization for multi-stage pipelines by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/1958
* [BugFIX] enable Hunyuan image3 with stage selection among text_to_image/image_to_text by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/1826
* [feature] stable_audio_open_1 teacache support by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/1314
* [Feature] Add a extra body param in create video api by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1955
* [Bugfix] Support base64 input for --ref-audio in Qwen3-TTS client by @lolyhop in https://github.com/vllm-project/vllm-omni/pull/1389
* [Feature] support to change the speaker of qwen3-omni by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/1963
* [skip ci] Keep the latest version. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/2112
* [Enhancement] Add force_refresh support for GLM-Image for cache-dit 1.3.0 upgrade by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1858
* [Bugfix] fix offload and hsdp incompatibility by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1888
* [Feature]: add Ulysses advanced_uaa mode by @dongbo910220 in https://github.com/vllm-project/vllm-omni/pull/1379
* [Test] L4 complete diffusion feature test for Qwen-Image-Layered models by @kechengliu97 in https://github.com/vllm-project/vllm-omni/pull/2085
* [Refactor] Unify torch profiler for omni and diffusion models by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/2099
* [API] Add layers and resolution parameters to /v1/images/edits endpoint by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/2053
* [CI] [RL]: Add e2e test for custom pipeline  by @knlnguyen1802 in https://github.com/vllm-project/vllm-omni/pull/2005
* [Perf] Qwen-Image Nightly Performance CI Improvement by @wtomin in https://github.com/vllm-project/vllm-omni/pull/2111
* [CI] Add conditions for L3 (tests after merging) and L4 (tests for nightly). by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1514
* [Enhancement] Custom chunk_size for mimo-audio model by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/1964
* [CI] Trigger nightly diffusion benchmark collects and html generates. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/1995
* [Core] Unified quantization framework by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/1764
* [Fix CI] Reduce num gpus to prevent ci failure by @wtomin in https://github.com/vllm-project/vllm-omni/pull/2131
* [Feat] Support scalar types in AdditionalInformationEntry by @NickCao in https://github.com/vllm-project/vllm-omni/pull/2105
* [Docs][skip ci] Fix omni and tts docs by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/2130
* [Bugfix] Fix high TTFP for Base task in Gradio TTS demo by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2116
* [Feature] Speech batch entrypoint by @divyanshsinghvi in https://github.com/vllm-project/vllm-omni/pull/1701
* [Bugfix] Fix memory leak: missing chunk_transfer_adapter.cleanup() in OmniARScheduler by @dubin555 in https://github.com/vllm-project/vllm-omni/pull/2028
* [Fix] Qwen3 TTS audio handling for long ref_audio by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/2104
* [CI/Build] Fix Doc 404s by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/2155
* [Voxtral TTS] Add multilingual support in gradio demo by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2151
* Add TTS Text Preprocessing to Gradio Demo by @rohinarora73 in https://github.com/vllm-project/vllm-omni/pull/2152
* [Docs] Update WeChat QR code for community support by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/2165
* [Voxtral TTS] Use 8 step flow matching instead of 16 by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2158
* [CI] Add online e2e test for qwen2.5 omni by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/1668
* [Test] Add L4 diffusion feature test for LongCat-Image by @lcukyfuture in https://github.com/vllm-project/vllm-omni/pull/1970
* [DOC] intel GPU model support list by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/2154
* [Test] L4 complete diffusion feature test for LongCat Image Edit models by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/2035
* [CI] Fix examples tests error by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/2138
* [Fixbug] increase qwen2 5 online test timeout limit by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/2171
* [Docs] refine UAA documentation by @dongbo910220 in https://github.com/vllm-project/vllm-omni/pull/2148
* [CI] Add Stable Diffusion 3.5 Tests by @spencerr221 in https://github.com/vllm-project/vllm-omni/pull/2120
* [Cleanup] Remove stray test file from engine directory by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2161
* [Bug-Fix]fix bug of empty prompt input by @Hu1Lcode in https://github.com/vllm-project/vllm-omni/pull/2041
* [BugFix] Make Stage Device Initialization Respect Visible GPUs by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/2025
* [Test] Add Qwen-tts test cases and unify the style of existing test cases by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/1911
* [Perf] [TTS] Improve Fish Speech S2 Pro voice cloning TTFP by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/2145
* Revert "[Test] Add Qwen-tts test cases and unify the style of existing test cases" by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2192
* [CI] Skip test_sd3_expansion due to CI failure 5148 by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/2191
* [Frontend] Speaker embedding support for speech and voices APIs by @marksverdhei in https://github.com/vllm-project/vllm-omni/pull/1227
* [Bugfix] add inject model_arch to hf_overrides by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/2178
* [CI] Add nightly-test label trigger. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/2172
* [Bugfix] resolve stage config for GLM-Image with diffusers format by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/1894
* [Bugfix] Maintain model-level CPU offload in a blocking way by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/1978
* [skip ci][Docs] Add FlashAttention requirement for audio generation to prevent noise-only outputs in mimo-audio model by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/2205
* [Test]Add FLUX.2-dev online serving expansion test by @yangjianjuan in https://github.com/vllm-project/vllm-omni/pull/2174
* [Bugfix] Fix qwen3-omni async thinker to talker decode alignment for #1758 by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/2019
* [Fix] [skip ci] Fix path. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/2204
* [Bugfix] remove default sampling parameters by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/2173
* [BugFix]Fix keyError: num_processed_tokens_delta by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/2213
* [Enhancement] Patch AsyncOmniEngine try_get_output[_async] hanging issues by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/2153
* [Accuracy Benchmark] feat: add accuracy benchmark integrations for t2i and i2i by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/1917
* [Test] L4 complete diffusion feature test for Wan2.2 models by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/2087
* [Bug Fix] GLM-Image stage device isolation and t2i prompt preprocessing in Omni runtime by @JaredforReal in https://github.com/vllm-project/vllm-omni/pull/2137
* [CI] qwen2.5-omni model cannot recognize the synthetic video by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/2211
* [Bugfix] Fix Voxtral TTS voice embeddings not loading by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2156
* [CI] fix Wan22 timeout and i2i accuracy threshold by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/2235
* [Qwen3TTS][ServingSpeech] Bugfix/voice upload and add optional ref_text by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/2046
* [Doc] Improve diffusion generation parameter docs for online serving by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/2051
* [Bugfix] Fix diffusion benchmark issues #1873 by @Dnoob in https://github.com/vllm-project/vllm-omni/pull/1897
* [Compatibility] Add Multiple Attention Backends Support in MIMO-Audio Tokenizer by @qibaoyuan in https://github.com/vllm-project/vllm-omni/pull/2225
* [Bug Fix] Resolve broken image issue when TP is enabled and no seed is provided. by @zhtmike in https://github.com/vllm-project/vllm-omni/pull/2176
* [Test] L4 complete diffusion feature test for Z-Image by @yinpeiqi in https://github.com/vllm-project/vllm-omni/pull/2132
* [CI] Increase diffusion initialization timeout from 600 to 700 seconds in online serving tests by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/2230
* [CI] Add Voxtral TTS e2e test by @y123456y78 in https://github.com/vllm-project/vllm-omni/pull/2023
* [Bugfix] Fix tp and Quantization incompatible for Flux by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/2180
* [CI] Skip tests due to L3 CI failure by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/2245
* Frontend] Support --dtype in qwen3_omni offline e2e script by @reidliu41 in https://github.com/vllm-project/vllm-omni/pull/2246
* [BugFix][Qwen3-Omni]Fixed the issue of incorrect answers for single words. by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/2239
* [Test] L4 complete diffusion feature test for Qwen-Image models by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/1869
* [Bugfix] Fix dynamic function call on collective_rpc of DiffusionWorker by @knlnguyen1802 in https://github.com/vllm-project/vllm-omni/pull/2217
* [Bugfix]fix_test_bagel_online by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/2237
* [CI] Add sd3 for test by @spencerr221 in https://github.com/vllm-project/vllm-omni/pull/2219
* [CI] Add online e2e test for MIMO-Audio by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/2129
* [CI] remove benchmark/testing comparison w/ other frameworks by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/2179
* [Feature] support sp for hunyuan by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/2163
* [Bugfix] Modify conftest.py set unspecified parameters by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/2263
* [Release] Upgrade NPU dockerfile & docs for v0.18.0 by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/2271
* [CI] Update pytest command to exclude specific test in nightly build by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/2272
* [bugfix] Remove duplicate yaml entry by @pi314ever in https://github.com/vllm-project/vllm-omni/pull/2279
* [Bugfix] Fix Fish Speech S2 Pro prompt handling for truncated audio & emotion tag by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/2268
* [Misc] Clean up unused diffusion timing args in examples by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/2266
* [Qwen3TTS][Bugfix] Replace vLLM fused layers with HF-compatible numerics in code predictor by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/2277

## New Contributors
* @lengrongfu made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1566
* @ZhanqiuHu made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1619
* @usberkeley made their first contribution in https://github.com/vllm-project/vllm-omni/pull/759
* @HonestDeng made their first contribution in https://github.com/vllm-project/vllm-omni/pull/336
* @jiangmengyu18 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1537
* @pi314ever made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1677
* @Lidang-Jiang made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1687
* @JackLeeHal made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1197
* @dubin555 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1614
* @RomanKoshkin made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1417
* @SHYuanBest made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1729
* @zhaotyer made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1201
* @oglok made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1337
* @NumberWan made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1465
* @Flink-ddd made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1582
* @ieaves made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1665
* @ahengljh made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1863
* @Fishermanykx made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1848
* @Joshna-Medisetty made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1934
* @DomBrown made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1985
* @Dnoob made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1956
* @Hu1Lcode made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2007
* @yjb767868009 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1470
* @y123456y78 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1803
* @patrickvonplaten made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2045
* @zJuuu made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2050
* @salmanmkc made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2070
* @meghaagr13 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1816
* @akshatvishu made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1314
* @lolyhop made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1389
* @NickCao made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2105
* @rohinarora73 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2152
* @lcukyfuture made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1970
* @spencerr221 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2120
* @yangjianjuan made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2174
* @bjf-frz made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2087
* @zhtmike made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2176
* @reidliu41 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/2246

**Full Changelog**: https://github.com/vllm-project/vllm-omni/compare/v0.16.0...v0.18.0