v0.24.0

## Highlights

This release features 285 commits from 112 contributors, including 40 new contributors.

vLLM-Omni `v0.24.0`  brings a large set of improvements across TTS, speech, diffusion, image generation, video generation, robot-policy serving, quantization, runtime architecture, request batching, cache behavior, CI, and hardware backend support.

This release expands production-ready multimodal serving coverage with new and improved models such as Higgs Audio V3, Qwen3-TTS, Ming-omni-tts, IndexTTS2, MOSS-TTS-local-v1.5, Step-Audio2, SDXL, Cosmos3, GR00T-N1.7, HunyuanImage3, DreamZero, Wan2.2 / Wan S2V, SenseNova-U1, JoyAI-VL-Interaction, and more. It also includes the vLLM 0.23 and 0.24 rebases, significant Omni stage runtime refactoring, diffusion request-level batching, async output materialization, improved streaming paths, and broad CUDA / ROCm / XPU / NPU coverage.

### Key Improvements

* **Aligned with the vLLM 0.24 release line**, including the vLLM 0.23.0 and v0.24.0 rebases, release-branch updates, and compatibility fixes across model loading, graph behavior, quantization, and serving paths. **(#4286, #4709)**
* **Expanded TTS and audio model support**, adding or improving Higgs Audio V3, Qwen3-TTS, CosyVoice, Fish Speech, VoxCPM2, Voxtral, Ming-omni-tts, IndexTTS2, MOSS-TTS-local-v1.5, Step-Audio2, soulx-singer, Aura, and Stable-Audio-Open. **(#4169, #4204, #3838, #4664, #4341, #464, #3862, #4257, #3664)**
* **Made speech serving more production-ready**, with SSE audio streaming, word-level timestamps, Qwen3-TTS correctness fixes, token accounting, WebSocket input handling, streaming-tail fixes, cross-request audio-corruption fixes, and better voice-clone reference-audio behavior. **(#4034, #4490, #4679, #4673, #4731, #4706, #4415, #4429)**
* **Strengthened diffusion, image, and video generation**, including major improvements for Cosmos3, HunyuanImage3, DreamZero, BAGEL, Wan2.2, Wan S2V, LTX-2.3, SDXL, Qwen-Image, Qwen-Image-Edit, MammothModa2, GLM-Image, and SenseNova-U1. **(#4331, #4379, #4514, #4627, #4041, #4213, #4098, #4667, #4739, #4474, #4293, #4299, #3956, #4188)**
* **Improved core Omni runtime and serving infrastructure**, including the Omni stage runtime and distributed replica control plane, orchestrator output-path optimization, async output materialization, diffusion request-level batching, HF-config-based pipeline resolution, structured config classes, and configurable video storage backends. **(#3855, #4527, #4476, #4079, #3760, #4425, #2531)**
* **Expanded quantization and memory-efficiency coverage**, including ModelOpt FP8 for video-generation models, Qwen3-Omni NVFP4 W4A4 on Blackwell, Qwen-Image AutoRound W4A16, ROCm MXFP4 online quantization, HSDP + FP8 compatibility, CacheDiT cleanup, prefix-cache fixes, and improved memory metrics. **(#3305, #4025, #4528, #3588, #4104, #4494, #2527, #4449, #4106, #4477)**
* **Improved hardware backend support**, with fixes and features for CUDA, ROCm, XPU, and Ascend NPU, including NPU graph behavior, custom-op registration, diffusion-worker initialization, XPU CI fixes, platform-agnostic device selection, and ROCm test stability. **(#4454, #4712, #4386, #4674, #4451, #4496, #4380, #4574, #4821, #4848)**

### Core Architecture & Runtime

* Refactored the Omni stage runtime and distributed replica control plane, providing a stronger foundation for multistage omni serving and distributed stage coordination. **(#3855)**
* Reduced orchestrator bottlenecks by separating inter-stage outputs from client-facing outputs. **(#4527)**
* Added async Omni output materialization so generated multimodal artifacts can be surfaced through more flexible serving flows. **(#4476)**
* Added request-level batching for diffusion pipelines, improving throughput and serving efficiency for compatible diffusion workloads. **(#4079)**
* Introduced HF-config-based pipeline resolution and eager pipeline registry behavior, improving automatic pipeline selection and startup behavior. **(#3760)**
* Added structured `VllmOmniConfig` classes and restored deploy-config-based pipeline resolution in `StageConfigFactory`. **(#4425, #4729)**
* Migrated existing pipelines to `SupportsComponentDiscovery` / `ModuleDiscovery`, improving offload discovery and nested HSDP DiT sharding support. **(#3076, #3456)**
* Extracted diffusion output formatting boundaries and removed dead custom stage input hooks to simplify stage-processing internals. **(#4407, #4531)**
* Fixed multi-replica stage identity and cleaned up redundant deploy configuration. **(#4410)**
* Aligned Omni request input signatures with vLLM and fixed multimodal generation-stage output construction. **(#4568, #4579)**

### Model Support

* Added support for **bosonai/higgs-audio-v3-tts-4b** and optimized Higgs Audio V3 serving paths. **(#4169, #4204)**
* Added or improved TTS model families including **Ming-omni-tts dense 0.5B**, **Ming-omni-tts MoE 16.8B-A3B**, **IndexTTS2**, **MOSS-TTS-local-v1.5**, **Step-Audio2**, **soulx-singer**, **Aura**, and fine-tuned Qwen3-TTS language support. **(#2906, #4341, #3838, #4664, #4804, #464, #3862, #4257, #4210)**
* Added or enabled diffusion and image/video model coverage for **SDXL**, **GR00T-N1.7**, **Cosmos3**, **HunyuanImage3**, **DreamZero**, **Wan2.2 / Wan S2V**, **LTX-2.3 I2V**, **JoyAI-VL-Interaction**, **MammothModa2**, and **SenseNova-U1**. **(#4331, #3798, #4266, #4041, #4154, #4276, #4458, #4739, #4575, #4299, #4188)**
* Added DROID policy server support for Cosmos3 OpenPI and GR00T-N1.7 OpenPI serving, expanding robot-policy serving coverage. **(#4282, #3798)**
* Migrated or refreshed examples for Cosmos3, Helios, Magi human, VACE, MammothModa2-Preview, AudioX, Qwen Image, Stable-Diffusion-3.5, Voxtral TTS, and Qwen3-TTS. **(#4581, #4569, #4572, #4648, #4691, #4607, #4236, #4052, #4051, #4026)**

### Audio, Speech & TTS Serving

* Improved Qwen3-TTS serving correctness across custom voice mode, WebSocket input handling, no-async-chunk payload splitting, multi-token replay on resume, seeded residual MTP sampling, token accounting, Code2Wav CUDA Graph output length, and full-payload completion behavior. **(#4650, #4731, #4870, #4559, #4869, #4889, #4673, #4525, #4465)**
* Added streaming TTS word-level timestamps via forced alignment. **(#4034)**
* Added SSE `stream_format` support for audio speech streaming and made speech streaming default to SSE. **(#4490, #4679)**
* Fixed Code2Wav async-chunk streaming tail drop, reducing boundary clicks and time-compression artifacts. **(#4706)**
* Improved CosyVoice and CosyVoice3 paths, including TensorRT optimization, reference-text instruction templating, and prompt handling fixes. **(#4168, #4756)**
* Fixed Fish Speech serving issues including Gradio default voice handling, prefix-cache collision, and KV-cache unpacking after vLLM compatibility changes. **(#3941, #4008, #4428)**
* Improved MOSS-TTS stability and performance with eager initialization, CUDA Graph support, cross-request audio-corruption fixes, talker micro-optimizations, loading shims, streaming support, and WER accuracy fixes. **(#3230, #4157, #4415, #4230, #4398, #4804, #4859)**
* Improved VoxCPM2 quality, hardware adaptation, KV-cache settings, unified decode graph optimization, and test stability. **(#4319, #4310, #4374, #4279, #4606, #4878)**
* Updated audio and TTS recipes, baselines, and docs for Qwen3-TTS, VoxCPM, Voxtral, Stable-Audio-Open, AudioX, and Ming-omni-tts. **(#4026, #4521, #4051, #4774, #3664, #4607, #4324)**

### Diffusion, Image & Video Generation

* Improved Cosmos3 generation, transfer, quality, distributed behavior, testing, and documentation, including v2v support, transfer paths, regression fixes, I2V conditioning optimization, skipped unused conditioning latents, regional compile, sequence-parallel sound latent padding, OpenPI serving, and L2 serving tests. **(#4266, #4379, #4514, #4627, #4467, #4614, #4485, #4678, #4282, #4535)**
* Improved HunyuanImage3 correctness and performance with DiT grouped step batching, more resolution support, AR RGB conversion alignment, CoT truncation fixes, streaming CoT display, sync removal, timestep scalar-sync avoidance, KV prefetch, MoE fixes, and bridge request batching fixes. **(#4041, #4004, #4502, #4260, #4148, #4401, #4363, #4448, #4893, #4849)**
* Advanced DreamZero serving with TP and cross-attention cache fixes, CUDA Graph support, `torch.compile`, DiT caching, and engine-level KV-cache management groundwork. **(#4154, #4213, #4534)**
* Improved BAGEL generation with denoising schedule fixes, batched CFG forward reimplementation, CFG-parallel mixin reuse, memory metrics, and SP denoise fixes. **(#4509, #4098, #4768, #4477, #4328)**
* Improved Wan2.2 / Wan S2V with graph-break fixes, CacheDiT + Ulysses fixes, S2V API support, SP and HSDP support, Wan VAE spatially sharded decode, and VACE cache / lifecycle fixes. **(#4053, #3927, #3394, #4276, #4458, #4620, #4667)**
* Added LTX-2.3 I2V support, VAE decode parallelism, resident auxiliary modules, and tensor-parallel gated-attention fixes. **(#4739, #4277, #4144, #4439)**
* Improved image-generation serving with file-style image `response_format`, default image-generation limits, Qwen-Image RoPE fixes, Qwen-Image-Edit performance fixes, Qwen-Image-Layered compatibility, and Ming-flash-omni image-generation batching fixes. **(#1673, #3381, #4474, #4293, #4777, #4837)**
* Improved SenseNova-U1 with LoRA support, CFG parallelism, fused RMSNorm + 3D RoPE kernels, TeaCache, and pipeline fixes. **(#3971, #4188, #4669, #4164, #4326)**

### Quantization, Cache & Memory Efficiency

* Added ModelOpt FP8 support for Wan2.2 and HunyuanVideo-1.5 video generation, and propagated quantization configuration into HunyuanVideo-1.5 I2V transformer layers. **(#3305, #4245)**
* Added Qwen3-Omni NVFP4 W4A4 serving on Blackwell and fixed Thinker `lm_head` prefix handling so NVFP4 exclude lists are honored. **(#4025, #4528)**
* Added Qwen-Image AutoRound W4A16 support and fixed Qwen2.5-Omni AutoRound loading. **(#3588, #4781)**
* Added ROCm support for MXFP4 W4A4 online quantization. **(#4104)**
* Improved HSDP + FP8 online quantization compatibility and gated quack FP8 GEMM to datacenter Blackwell GPUs. **(#4494, #4817)**
* Simplified CacheDiT integration and improved cache behavior across DFlash, Qwen3-TTS, Fish Speech, DreamZero, Wan2.2, and OmniTensorPrefix. **(#2527, #4449, #3689, #4317, #4008, #4534, #3927, #4106)**
* Added memory metrics fixes and reporting improvements for BAGEL, Qwen-Image, non-text requests, and audio RTF metrics. **(#4477, #4517, #4482, #4794)**

### Serving, Frontend & API Behavior

* Added configurable video storage backends with TTL. **(#2531)**
* Returned proper HTTP status codes for audio voice endpoints and refactored guardrail error handling with 400 error codes. **(#3969, #4297)**
* Added async output materialization and streaming diffusion video generation output. **(#4476, #3737)**
* Added support for image-generation responses with `response_format: file`. **(#1673)**
* Fixed request batching regressions and output construction issues across diffusion and multimodal serving paths. **(#4079, #4579, #4837, #4910)**
* Added support for pipeline resolution based on HF config and eager pipeline registry initialization. **(#3760)**
* Replaced hardcoded CUDA device selection with platform-agnostic APIs. **(#4496)**
* Reverted the MediaConnector image/video URL-fetching change after follow-up validation needs. **(#4751)**

### Platforms, Distributed Execution & Hardware Coverage

* Improved Ascend NPU support with VoxCPM2 support, Qwen3-TTS 310P support, model-runner fixes, `OmniMRotaryEmbedding` support, diffusion-worker Ascend config initialization, custom-op registration, graph behavior restoration, HunyuanImage3 MoE EP+SP mapping, and v0.24.0 upgrade work. **(#4310, #4283, #4454, #3609, #4386, #4712, #4674, #4344, #4848)**
* Improved ROCm stability with CI grouping, diffusion-test fixes, Voxtral fixes, AMD CI updates after the v0.24.0 rebase, and MXFP4 online quantization support. **(#4208, #4574, #4380, #4821, #4104)**
* Improved XPU coverage with sage-attn backend support, Cosmos3 CUDA-hardcode removal, DreamZero any-hardware support, docker-build improvements, marker fixes, SDXL enablement, and v0.24 CI updates. **(#3785, #4360, #4399, #4402, #4451, #4331, #4805)**
* Improved distributed diffusion paths with HSDP/SP fixes and features across Cosmos3, Wan S2V, Wan VAE, nested DiT sharding, and CFG-parallel serving. **(#4452, #4678, #4276, #4458, #4620, #3456, #4900)**

### CI, Testing, Benchmarks & Documentation

* Improved ready/merge, rebase, and nightly pipelines with diff-aware L2/L3 upload, full E2E rebase coverage, nightly L2/L3 E2E groups, split diffusion X2V tests, Cosmos3 L2 tests, and reduced ready-CI duration. **(#4291, #4313, #4365, #4478, #4532, #4693, #4734, #4744, #4535, #4354)**
* Added reliability and stability tests for HunyuanImage3 and VoxCPM2, plus targeted skip/unskip changes for unstable or OOM-prone test cases. **(#4600, #4311, #4701, #4758, #4759, #4391)**
* Updated performance baselines for VoxCPM, Qwen3-TTS, Cosmos3, and perf JSONs. **(#4289, #4521)**
* Added and refined Claude / agent skills for precheck, quantization, vLLM-Omni test generation, and diffusion-model contribution workflows. **(#4216, #4252, #4434, #4697, #4110)**
* Updated docs and recipes for supported models, Qwen Image, Qwen3-TTS, Voxtral TTS, Stable-Diffusion-3.5, Stable-Audio-Open, Cosmos3 FP8, CUDA custom Docker builds, failure modes, and release metadata. **(#4233, #4300, #4236, #4026, #4051, #4774, #4052, #3664, #4393, #4584, #1386, #3926)**

## What's Changed
* [TTS][New Model] support bosonai/higgs-audio-v3-tts-4b by @yuekaizhang in https://github.com/vllm-project/vllm-omni/pull/4169
* [Perf][Higgs-Audio-V3] LRU cache for voice-clone ref-audio encode by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4200
* [Perf][Higgs-Audio-V3] Turn on Stage-0 prefix caching by default by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4199
* [Doc] Add Stable-Diffusion-3.5 recipe for 1x RTX A6000 48GB (#2645) by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4052
* [Bugfix] Fish Speech Gradio hardcoded default voice causes 400 error by @nagelanping in https://github.com/vllm-project/vllm-omni/pull/3941
* [Model] Support languages added by fine-tuned Qwen3-TTS checkpoints by @n0n4m39911 in https://github.com/vllm-project/vllm-omni/pull/4210
* [Quant] ModelOpt FP8 for Wan2.2 & HunyuanVideo-1.5 video-gen by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/3305
* [BugFix] fix hunyuan image3 offline cot by @BLANKETusers in https://github.com/vllm-project/vllm-omni/pull/4174
* [skip ci] docs: update WeChat QR code by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4242
* [Refactor]Refactoring audio_in_video implementation by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/3566
* [Doc] Add precheck-pr Claude Code skill by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4216
* [Perf] Keep LTX2.3 auxiliary modules resident by default by @mglyn in https://github.com/vllm-project/vllm-omni/pull/4144
* [Fix][HiggsAudioV3] Fix ramp-down off-by-one crash and buffer state bugs in Stage0 talker by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4219
* Cosmos3 video to video generation by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4266
* [Quant][Perf] Default Blackwell FP8 GEMM to quack CuteDSL fused-bias kernel by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/4241
* [Refactor] Remove dead legacy pipeline.yaml loader and duplicate diff… by @AbelSara in https://github.com/vllm-project/vllm-omni/pull/4023
* [Core] Cleanup DiffusersPipelineLoader by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/1932
* docs: update README and supported models for v0.22.0 release by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4233
* fix: Propagate Quantization Configuration to HunyuanVideo-1.5 I2V Transformer to Enable FP8 Layers by @weizhoublue in https://github.com/vllm-project/vllm-omni/pull/4245
* [Perf] lance perf optimize (t2i & i2i) by @yangjianjuan in https://github.com/vllm-project/vllm-omni/pull/4214
* [Bugfix] Register LTX RMSNorm identity weight as buffer by @mglyn in https://github.com/vllm-project/vllm-omni/pull/4278
* Fix README.md typo by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/4292
* [XPU] Add sage_attn backend by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/3785
* [Feat] Add vae-decode-parallel for LTX-2.3 by @mglyn in https://github.com/vllm-project/vllm-omni/pull/4277
* [ROCm] [CI] Add group feature and envs feature by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/4208
* [Feature] LoRA support for SenseNova-U1 by @leohuang257 in https://github.com/vllm-project/vllm-omni/pull/3971
* [BugFix] Fix HunyuanImage3 CoT truncation when stop token stripped by detokenizer by @zengchuang-hw in https://github.com/vllm-project/vllm-omni/pull/4260
* [Perf][Bugfix] qwen3-tts hot path: prefix-cache OOM guards + talker/orchestrator micro-opts by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/3689
* add moriio transfer engine by @inkcherry in https://github.com/vllm-project/vllm-omni/pull/1742
* [skip ci] fix(config): pin VoxCPM2 KV cache to avoid OOM on small GPUs by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4279
* [Bugfix] purge chunk-transfer zombies on every schedule tick to keep engine-core alive on aborts (fixes #3736) by @abinggo in https://github.com/vllm-project/vllm-omni/pull/3774
* [BugFix] Remove pydub dependency for Python 3.13 compatibility by @FED4 in https://github.com/vllm-project/vllm-omni/pull/4035
* [CI] skip unrelated L3 merge tests via diff-aware upload by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4291
* [Perf][TTS] Optimize cosyvoice TTFP and throughput using TensorRT by @yuekaizhang in https://github.com/vllm-project/vllm-omni/pull/4168
* [Bugfix] Fix Fish Speech prefix cache collision from missing cache_salt by @nagelanping in https://github.com/vllm-project/vllm-omni/pull/4008
* [Test] skip oom test case for issue #4285 by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/4311
* [Docs] Update README supported models section with TTS and Diffusion categories by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4300
* [Model] Add Ming-omni-tts dense 0.5B pipeline by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/2906
* [CI/Build] Voxtral TTS Tests by @clodaghwalsh17 in https://github.com/vllm-project/vllm-omni/pull/3738
* [BugFix] moss_tts_nano: eager-init lm + audio_tokenizer in __init__ so load_format: dummy works by @leohuang257 in https://github.com/vllm-project/vllm-omni/pull/3230
* [skip ci] Add width and height args to offline i2i example script by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/4031
* [Bugfix] qwen3-tts prefix cache: drop per-key size cap that corrupted… by @JuanPZuluaga in https://github.com/vllm-project/vllm-omni/pull/4317
* Step audio R1 reasoning parser by @QiuMike in https://github.com/vllm-project/vllm-omni/pull/2846
* [Test] Automatically clean up audio files generated from requests, and realtime invalid-param coverage by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4294
* [Skills] Add quantization Claude skill by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4252
* [Docs] add doc for failure mode by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/3926
* [Bugfix][HunyuanImage] fix accuracy in stream mode by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/4265
* [Bugfix][VoxCPM2] Fix VoxCPM2 concurrent speech quality  by @Shirley125 in https://github.com/vllm-project/vllm-omni/pull/4319
* [Bugfix] Fix SP denoise indentation bug on BAGEL by @kushanam in https://github.com/vllm-project/vllm-omni/pull/4328
* [CI] skip unrelated L2 ready tests via diff-aware upload, aligned L3 tweak and fix issue 4334 by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4313
* [NPU] Support VoxCPM2 model by @tanhaoan333 in https://github.com/vllm-project/vllm-omni/pull/4310
* [Doc] Clean up PR template by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4336
* [Bugfix] Add Cosmos3-Nano baselines and fix USP gather by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4301
* [WAN2.2-S2V] Add server API for image + audio by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/3394
* [HunyuanImage][Perf] opt prepare_attention_mask for e2e latency 6% reduction by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/4333
* [Perf] Optimize Higgs Audio v3 serving by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4204
* [refactor] Refactor guardrail error handling - add 400 error code by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4297
* Feat: `non_streaming_mode` for Qwen3-TTS Base Models During Online Inference by @nagisa-kunhah in https://github.com/vllm-project/vllm-omni/pull/4198
* feat(moss-tts): add CUDA Graph support for codec decoder by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4157
* [Bugfix] Fix Qwen3-TTS gradio streaming TTFP by using audio/pcm by @yuxinyuan in https://github.com/vllm-project/vllm-omni/pull/4346
* [XPU][COSMOS3] removing cuda hardcode and make VLLM_VIDEO_SYNC_TIMEOUT a tunable config by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4360
* [Model] Add more resolution support for HunyuanImage3.0 by @Semmer2 in https://github.com/vllm-project/vllm-omni/pull/4004
* [Refactor] Output Processor Phase 2: separate multimodal output channel (#1601) by @meghaagr13 in https://github.com/vllm-project/vllm-omni/pull/2744
* [platform] fix: set UnspecifiedOmniPlatform device_type to cpu by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/4357
* [bugfix]VoxCPM2 audio encoder adapt other than CUDA by @tanhaoan333 in https://github.com/vllm-project/vllm-omni/pull/4374
* [ci]skip voxcpm2 pcm hnr test  by @Shirley125 in https://github.com/vllm-project/vllm-omni/pull/4375
* [Hardware][Ascend] Adapt Qwen3 TTS for 310P by @zyz111222 in https://github.com/vllm-project/vllm-omni/pull/4283
* [CI] Diff-gate L2/L3 E2E jobs and migrate single-GPU tests to gpu_1_queue by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4365
* [CI] skip MOSS-TTS-Nano E2E tests pending issue#4361 by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4391
* [Perf][Wan2.2] Skip attention mask for zero-padded SP sequences to avoid varlen path by @jjuvonen-amd in https://github.com/vllm-project/vllm-omni/pull/3763
* [Bugfix] Fix cross-request codes.ref leak in Qwen3-TTS make_omni_output by @henryj in https://github.com/vllm-project/vllm-omni/pull/4373
* [New Model] Step-Audio2  by @wuli666 in https://github.com/vllm-project/vllm-omni/pull/464
* [Bugfix] Ming-omni-tts: generalize dict checks to Mapping by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4397
* [XPU][CI] Fix docker build slowness by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4402
* [Refactor] TTS serving adapter framework + migrate TTS models (excl. ming_flash_omni) by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4330
* [Bugfix] Surface all-rank diffusion RPC failures by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4403
* [Doc][Recipe] Update CUDA verifications for inclusionAI/Ming-omni-tts-0.5B by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4324
* [XPU] update Dreamzero to support any HW by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4399
* [Rebase] Rebase to vllm 0.23.0 by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/4286
* [BugFix] Fix HunyuanImage3 size match issue by @Semmer2 in https://github.com/vllm-project/vllm-omni/pull/4416
* [Perf] Restore parallel stage initialization for AR+DiT pipelines by @zengchuang-hw in https://github.com/vllm-project/vllm-omni/pull/3641
* [Doc] validate Cosmos3-Nano online FP8 by @leohuang257 in https://github.com/vllm-project/vllm-omni/pull/4393
* [Perf] change omniconnector info log to debug log to avoid mass printing by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/4426
* Update WeChat QR code by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4431
* [Bugfix] Qwen3-TTS: trim reference audio in no_async_chunk voice clone by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4429
* [Bugfix] Fix LTX-2.3 tensor-parallel gated attention by @mglyn in https://github.com/vllm-project/vllm-omni/pull/4439
* [Feature] Add HunyuanImage3 DiT grouped step batching by @TaffyOfficial in https://github.com/vllm-project/vllm-omni/pull/4041
* Support DROID policy server for Cosmos3 OpenPI by @yuzhudong in https://github.com/vllm-project/vllm-omni/pull/4282
* [CI/Build] Skip L2/L3 CI for pytest skip-mark changes by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4422
* [Refactor] Extract diffusion output formatting boundary by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4407
* [Perf][HunyuanImage] remove sync logic by @Bounty-hunter in https://github.com/vllm-project/vllm-omni/pull/4401
* [XPU][CI] Fix ERROR ...test_glm_image_sp.py - Failed: 'sp' not found in markers  by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4451
* [Enhancement] Introduce High-Performance MoT (Mixture-of-Tokens) Kernels: Triton Implementation by @princepride in https://github.com/vllm-project/vllm-omni/pull/3960
* [Bugfix] Cosmos3: fix HSDP dtype mismatch on time_embedder via _hsdp_ignored_modules by @KwokhoTsui in https://github.com/vllm-project/vllm-omni/pull/4452
* [BugFix] higgs-audio-v3: load codec lazily under load_format=dummy by @yuekaizhang in https://github.com/vllm-project/vllm-omni/pull/4368
* [Bugfix] Fix multi-replica stage identity and clean up redundant deploy config by @ZhengWG in https://github.com/vllm-project/vllm-omni/pull/4410
* fix(models): remove invalid sampling_metadata parameter from logits_processor by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4456
* [Bugfix] [CI] [ROCm] [CUDA] Fix voxtral test by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/4380
* [CI][Rebase]: run full e2e suite in rebase Ready/Merge pipelines in the rebase pipeline[skip-ci] by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/4478
* [BugFix][NPU]: Fix npu model runner too many values to unpack ($[1][0][4]) by @FayeSpica in https://github.com/vllm-project/vllm-omni/pull/4454
* [BugFix][Qwen-Image] align txt_seq_lens RoPE width with padded embeds by @SamitHuang in https://github.com/vllm-project/vllm-omni/pull/4474
* [Bugfix] Fix DFlash prefix cache corruption due to missing lookahead block by @NickCao in https://github.com/vllm-project/vllm-omni/pull/4449
* [SDXL] SDXL model enabling by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4331
* [Perf][qwen-image-edit] Skip attention-mask to avoid varlen path by @kTorp in https://github.com/vllm-project/vllm-omni/pull/4293
* [fix bug] add default limit for image generation by @LJH-LBJ in https://github.com/vllm-project/vllm-omni/pull/3381
* [Doc] Add guides for custom docker image build on NVIDIA CUDA [Skip-CI] by @loveysuby in https://github.com/vllm-project/vllm-omni/pull/1386
* [BugFix] Fix the issue of incorrect generated tokens by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4461
* [CI] Update perf json baselines by @congw729 in https://github.com/vllm-project/vllm-omni/pull/4289
* [Bugfix] Fix missing memory metrics info for bagel (ar->dit diffusion) by @natureofnature in https://github.com/vllm-project/vllm-omni/pull/4477
* [Refactor] Omni Stage Runtime and Distributed Replica Control Plane by @yinpeiqi in https://github.com/vllm-project/vllm-omni/pull/3855
* [Wan2.2] Fix "uninitialized nn.Module of type RotaryEmbeddingWan" graph-break by @kTorp in https://github.com/vllm-project/vllm-omni/pull/4053
* [Perf][PrefixCache] Avoid per-step blocking write in OmniTensorPrefix… by @LHXuuu in https://github.com/vllm-project/vllm-omni/pull/4106
* [Doc] Qwen image 2512 recipe by @AbelSara in https://github.com/vllm-project/vllm-omni/pull/4236
* [bugfix] Shim PreTrainedModel._tp_plan to fix MOSS-TTS-Nano load crash by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4398
* [Docs] Fix hallucinated stage CLI flags in documentation and comments  by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4512
* [Bugfix] Wire log_stats to AsyncOmni and add missing token metrics for non-text requests by @blondeCS in https://github.com/vllm-project/vllm-omni/pull/4482
* [Refactor] Migrate existing pipelines to use SupportsComponentDiscovery for offload discovery by @NickCao in https://github.com/vllm-project/vllm-omni/pull/3076
* Refactor: extract OmniStreamingVideoHandler base and QwenOmniStreamingVideoHandler by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/4424
* [DEBUG]Add has_preprocess branch to _dummy_run for models with preprocess stage by @Wallbreazzz in https://github.com/vllm-project/vllm-omni/pull/4189
* [Quantization][Qwen3-Omni] Enable NVFP4 W4A4 serving on Blackwell by @YIHONG-JIN in https://github.com/vllm-project/vllm-omni/pull/4025
* [BugFix]: Fix Bagel generate_image denoising schedule by @princepride in https://github.com/vllm-project/vllm-omni/pull/4509
* Remove unregister_vllm_metrics no-op patch by @vraiti in https://github.com/vllm-project/vllm-omni/pull/4396
* [Perf]Enable regional compile for Cosmos3 HSDP blocks by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/4485
* Update VoxCPM and Qwen3TTS perf baseline. by @congw729 in https://github.com/vllm-project/vllm-omni/pull/4521
* Add NPU support for OmniMRotaryEmbedding by @Wallbreazzz in https://github.com/vllm-project/vllm-omni/pull/3609
* [SKILLS] Add vllm-omni-test agent skill for CI-aligned test generation by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4434
* [Refactor]Base class for dit pipelines with unified parameter declaration[1/N] by @princepride in https://github.com/vllm-project/vllm-omni/pull/4225
* [Doc] Qwen3-TTS: add 0.6B on 1x RTX 4090 and fix broken links by @zeningc in https://github.com/vllm-project/vllm-omni/pull/4026
* [CI] skip ci for issue 4537 by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/4538
* [Feat] DreamZero fix tp & cross attn cache by @AbelSara in https://github.com/vllm-project/vllm-omni/pull/4154
* [Bugfix] MOSS-TTS: fix cross-request audio corruption under batching by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4415
* [Doc] Add 1x RTX A6000 48GB section to Voxtral TTS recipe (#2645) by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4051
* [Cleanup] Remove duplicate HiDreamI1ImagePipeline (fixes #4009) by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4045
* [Perf] Stacked audio ops: batched matmul and embedding gather in MOSS-TTS talker by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4230
* [Perf][DreamZero]dreamzero supports CUDA graph, torch.compile, and DiT Caching. by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4213
* [Chore][TTS] Reference recipes.vllm.ai recipes, remove dead code, and fix stale references by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4567
* [Bugfix] Fix HSDP + FP8 online quantization compatibility by @KwokhoTsui in https://github.com/vllm-project/vllm-omni/pull/4494
* [Frontend] Configurable video storage backends with TTL by @ieaves in https://github.com/vllm-project/vllm-omni/pull/2531
* [Cache Refactor 1/N] Simplify CacheDiT Integration by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/2527
* [ROCm] [CI] Fix diffusion test by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/4574
* [Diffusion] add GR00T-N1.7 pipeline with OpenPI serving by @timzsu in https://github.com/vllm-project/vllm-omni/pull/3798
* [Bugfix] Align vLLM Omni Request input signature with vLLM.  by @timzsu in https://github.com/vllm-project/vllm-omni/pull/4568
* [Bugfix][Qwen3-Omni] Fix Code2Wav CUDA-graph output length surplus (#4466) by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4525
* Cosmos3 transfer by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4379
* [Feature]Migrate helios example by @princepride in https://github.com/vllm-project/vllm-omni/pull/4569
* [Feature] JoyAI-VL-Interaction streaming interaction serving layer by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/4575
* [BugFix]Fix diffusion profile RPC None results by @yixiaoer in https://github.com/vllm-project/vllm-omni/pull/4372
* [Model] Aura support - Non async chunk path by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/4257
* Improve Cosmos3 v2v quality by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4514
* [Perf/Fix] Reimplement Batched CFG Forward for Bagel by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/4098
* [Bugfix] Fix Fish Speech kvcache attention KV cache unpacking for vLLM 0.23.0 by @nagelanping in https://github.com/vllm-project/vllm-omni/pull/4428
* [Cleanup] Remove dead/stale modules with no repo references (fixes #4009) by @yangyonggit in https://github.com/vllm-project/vllm-omni/pull/4044
* [Perf][Cosmos3] Optimize I2V conditioning transfer (8.63% faster x-inference) by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4467
* [chore][skills]: tightens the add-diffusion-model skill under .claude/skills to better match the current vllm-omni codebase and repo conventions by @RuixiangMa in https://github.com/vllm-project/vllm-omni/pull/4110
* [Doc] validate Cosmos3-Super online FP8 by @KwokhoTsui in https://github.com/vllm-project/vllm-omni/pull/4584
* Update WeChat group QR code by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4601
* [Bugfix][Qwen3-Omni] Pass Thinker lm_head prefix so NVFP4 exclude-list is honored by @YIHONG-JIN in https://github.com/vllm-project/vllm-omni/pull/4528
* [Feat] Streaming TTS word-level timestamps via forced aligner by @wjinxu in https://github.com/vllm-project/vllm-omni/pull/4034
* Add SSE stream_format support for audio speech streaming by @syd520zy in https://github.com/vllm-project/vllm-omni/pull/4490
* [Feature]: Migrate magi human example by @princepride in https://github.com/vllm-project/vllm-omni/pull/4572
* [Perf]perf: skip unused Cosmos3 I2V conditioning latents by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/4614
* [CI][NPU] Update VLLM ASCEND IMAGE by @FayeSpica in https://github.com/vllm-project/vllm-omni/pull/4602
* [CI][Rebase] Fix Ready/Merge pipeline upload by stripping source_file_dependencies by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/4532
* [Feature] Support Async Omni output materialization by @fake0fan in https://github.com/vllm-project/vllm-omni/pull/4476
* [Bugfix][HunyuanImage3] Align AR RGB conversion with official semantics by @TaffyOfficial in https://github.com/vllm-project/vllm-omni/pull/4502
* [Model] Add IndexTTS2 text-to-speech support by @BeatSeat in https://github.com/vllm-project/vllm-omni/pull/3838
* fix: return proper HTTP status codes for audio voice endpoints by @ieaves in https://github.com/vllm-project/vllm-omni/pull/3969
* [NFC] Update misleading seed-tts-eval optional dependency guidance by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4619
* Qwen3-TTS full-payload: emit empty-finished payload on degenerate take instead of dropping (fixes Stage-1 300s hang) by @henryj in https://github.com/vllm-project/vllm-omni/pull/4465
* [Doc] Stability AI/Stable-Audio-Open by @Ronnie-Rui in https://github.com/vllm-project/vllm-omni/pull/3664
* [BugFix] Use ModuleDiscovery for HSDP DiT sharding to support nested pipelines by @NickCao in https://github.com/vllm-project/vllm-omni/pull/3456
* [NPU] Initialize Ascend config for diffusion workers by @Fishermanykx in https://github.com/vllm-project/vllm-omni/pull/4386
* [Refactor] Replace hardcoded CUDA device selection with platform-agnostic API by @NickCao in https://github.com/vllm-project/vllm-omni/pull/4496
* [Refactor] Migrate cosmos3 example to standard task example + model_extra by @leohuang257 in https://github.com/vllm-project/vllm-omni/pull/4581
* [Bugfix] Fix make_request_output TypeError on multimodal generation-stage outputs by @QianCyrus in https://github.com/vllm-project/vllm-omni/pull/4579
* [feat] Streaming diffusion video generation output by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/3737
* Add field IDs to bug report issue template by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4605
* [Fix] JoyVL serving: align with reference engine (bounded long-term memory, timestamps, max_pixels) + recipe updates by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/4623
* Fixed Cosmos3 regression by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4627
* [Feature] add a new response_format is file type when image generate by @lengrongfu in https://github.com/vllm-project/vllm-omni/pull/1673
* [BugFix] Delete unnecessary log printing by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4566
* [Feat] Support Pipeline Resolution Based on HF Config & Make Pipeline Registry Eager by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/3760
* [perf] avoid HunyuanImage3 timestep scalar sync by @TaffyOfficial in https://github.com/vllm-project/vllm-omni/pull/4363
* [Model] add soulx-singer support by @MrDongsls in https://github.com/vllm-project/vllm-omni/pull/3862
* [S2V] Add sequence parallelism support to WanS2VTransformer3DModel by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4276
* [refactor] Flatten Ming-flash-omni-2.0 imagegen args by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4587
* [Bugfix] [AURA] Correct aura results when use qwen3-tts in custom voice mode by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/4650
* [Refactor] Migrate VACE example to standard task examples + model_extras by @ForestWisdom in https://github.com/vllm-project/vllm-omni/pull/4648
* [Feat] Enable streaming CoT display for HunyuanImage-3.0 AR generation by @zengchuang-hw in https://github.com/vllm-project/vllm-omni/pull/4148
* [Bugfix] Use MediaConnector for image/video URL fetching to prevent SSRF by @NickCao in https://github.com/vllm-project/vllm-omni/pull/2565
* [Test] Skip MOSS TTS offline E2E tests (issue #4700) by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4701
* [Bugfix][CI]Fix HSDP regional compile boundary by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/4668
* [BugFix] Fix Wan2.2-VACE-Fun cache and serving lifecycle by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4667
* [Test] add reliability test case for HunyuanImage-3.0-Instruct and reliability/stability test for VoxCPM2 by @zhumingjue138 in https://github.com/vllm-project/vllm-omni/pull/4600
* [BugFix] Fix MammothModa2 Text-to-Image Bug by @menjiantong in https://github.com/vllm-project/vllm-omni/pull/4299
* [Bugfix] Fix CFG-parallel broadcast dtype mismatch in GLM-Image by @dsocek in https://github.com/vllm-project/vllm-omni/pull/3956
* [BugFix] Remove prefix-cache test cases by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4703
* [Bugfix] Unified import of PretrainedConfig by @R2-Y in https://github.com/vllm-project/vllm-omni/pull/4692 
* [CI] Run L2/L3 E2E group on main nightly builds by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4693
* [Model] Support MOSS-TTS-local-v1.5 by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/4664
* [Config] Add structured VllmOmniConfig classes by @Acerak01-fy in https://github.com/vllm-project/vllm-omni/pull/4425
* [Bugfix][NPU] Register vllm-ascend custom ops in NPUOmniPlatform.set_device by @FayeSpica in https://github.com/vllm-project/vllm-omni/pull/4712
* [Core] hunyuan-image prefetch kv by @AbelSara in https://github.com/vllm-project/vllm-omni/pull/4448
* [Model] Add Ming-omni-tts MoE 16.8B-A3B + CFM CUDAGraph by @LHXuuu in https://github.com/vllm-project/vllm-omni/pull/4341
* [skip ci][Misc] LVSA showcase (training-free block-sparse attention) by @gglorian in https://github.com/vllm-project/vllm-omni/pull/4192
* Wan2.2: Fix the bug of using cache-dit with ulysses in t2v and i2v by @mengker33 in https://github.com/vllm-project/vllm-omni/pull/3927
* [Bugfix] Add back resolving from deploy config written pipeline in StageConfigFactory by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4729
* [Bugfix] Stream Qwen3-TTS WebSocket input as one request by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4731
* [Doc] Add code-quality dimension to precheck-pr skill by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4697
* [CI][Bugfix] Fix nightly L2/L3 E2E-only upload and Buildkite EXTRA escaping by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4734
* Revert "[Bugfix] Use MediaConnector for image/video URL fetching to prevent SSRF" by @Gaohan123 in https://github.com/vllm-project/vllm-omni/pull/4751
* [Bugfix] Fix /v1/audio/speech usage token accounting for Qwen3-TTS (#4646) by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4673
* Updated Cosmos3 docstrings by @MaciejBalaNV in https://github.com/vllm-project/vllm-omni/pull/4727
* [Feature] Spatially-sharded (SP) decode for the Wan VAE by @rahul-steiger-nv in https://github.com/vllm-project/vllm-omni/pull/4620
* [Test] Skip Qwen3-TTS batch E2E tests by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4758
* [Test] Un-skip Qwen3-TTS batch E2E; match documented omit-null response shape (#4757) by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4759
* [Bugfix] CosyVoice3: wrap ref_text in instruction template (#4644) by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4756
* [WAN_S2V] Enable HSDP for wan s2v by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4458
* [CI][Bugfix]Split nightly Diffusion X2V function tests by T2V/I2V to fix timeout by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4744
* [Refactor] Reuse CFGParallelMixin in Bagel for CFG-parallel denoising by @suyanli220 in https://github.com/vllm-project/vllm-omni/pull/4768
* [CI]Add Cosmos3 L2 serving tests by @bjf-frz in https://github.com/vllm-project/vllm-omni/pull/4535
* [Core] Remove dead custom_process_input_func hooks in stage input processors by @Nughm3 in https://github.com/vllm-project/vllm-omni/pull/4531
* Update WeChat QR code asset by @david6666666 in https://github.com/vllm-project/vllm-omni/pull/4778
* [CI] Reducing mithril-h100-pool in ready&merge CI, optimize ready CI duration by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4354
* Fix Qwen2.5-Omni AutoRound loading by @ayaka14732 in https://github.com/vllm-project/vllm-omni/pull/4781
* [BugFix] Restore pre-#9572 NPU graph behavior (cap cudagraph_mode to PIECEWISE) by @FayeSpica in https://github.com/vllm-project/vllm-omni/pull/4674
* [Core] Performance fix for orchestrator bottleneck: separates inter-stage from client outputs by @yinpeiqi in https://github.com/vllm-project/vllm-omni/pull/4527
* [Core][Frontend] Support request-level batching for diffusion pipelines by @yJader in https://github.com/vllm-project/vllm-omni/pull/4079
* Make speech streaming default to SSE by @syd520zy in https://github.com/vllm-project/vllm-omni/pull/4679
* [Model] Migrate MammothModa2-Preview example to standard task example + model_extras by @moisf56 in https://github.com/vllm-project/vllm-omni/pull/4691
* Migrate audiox examples by @zzehli in https://github.com/vllm-project/vllm-omni/pull/4607
* [Feat] Support CFG-Parallel For Sensenova-U1-8B-MoT by @sphinxkkkbc in https://github.com/vllm-project/vllm-omni/pull/4188
* [Quant][AutoRound] Add W4A16 support for Qwen-Image by @yiliu30 in https://github.com/vllm-project/vllm-omni/pull/3588
* [Perf]: Fused RMSNorm+3D Rope kernel in Sensenova-U1 by @sphinxkkkbc in https://github.com/vllm-project/vllm-omni/pull/4669
* [BugFix][Cosmos3] Pad sound latents so video+sound runs under sequence parallelism by @lishunyang12 in https://github.com/vllm-project/vllm-omni/pull/4678
* [Feat] Support teacache for SenseNova-U1 by @fywc in https://github.com/vllm-project/vllm-omni/pull/4164
* [Rebase] Rebase to vllm v0.24.0 by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/4709
* [Doc] Add Voxtral TTS RTX 5090 recipe by @ayaka14732 in https://github.com/vllm-project/vllm-omni/pull/4774
* [XPU CI][0.24] fix xpu ci by update pipeline-intel.yaml and triton version by @xuechendi in https://github.com/vllm-project/vllm-omni/pull/4805
* [ROCm] [CI] Fix AMD CI after v0.24.0 rebase by @tjtanaa in https://github.com/vllm-project/vllm-omni/pull/4821
* [codex] Update maintainers and code owners by @hsliuustc0106 in https://github.com/vllm-project/vllm-omni/pull/4823
* [Bugfix][Quant] Gate quack FP8 to datacenter Blackwell (sm_100.x) only by @wangrzneu in https://github.com/vllm-project/vllm-omni/pull/4817
* [Fix]: Set .transformer on SenseNova pipeline, warn if no .transformer in CacheDiT summary by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/4326
* [Feature] Improve diffusion LoRA compatibility for SDXL by @tthakkal in https://github.com/vllm-project/vllm-omni/pull/4825
* [BugFix] Fix Code2Wav async_chunk streaming tail drop (boundary clicks + ~1.2% time compression) by @aporix in https://github.com/vllm-project/vllm-omni/pull/4706
* [ci]loose voxcpm2 HNR baseline by @Shirley125 in https://github.com/vllm-project/vllm-omni/pull/4801
* [CI] Lower Qwen-Image e2e SSIM threshold from 0.97 to 0.94 by @NumberWan in https://github.com/vllm-project/vllm-omni/pull/4613
* [Bugfix] clean per-stage audio RTF and align audio metrics by @ZacheryAU in https://github.com/vllm-project/vllm-omni/pull/4794
* [NPU] Fix HunyuanImage3 MoE EP+SP parallel mapping by @Celeste-jq in https://github.com/vllm-project/vllm-omni/pull/4344
* [Bugfix] Qwen3-TTS talker handles multi-token replay on resume by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4559
* [Bugfix] Realign Serving Tokenization Name to Fix Error Handling Routes by @alex-jw-brooks in https://github.com/vllm-project/vllm-omni/pull/4816
* [Chore]: refactor out unused/redundant params in diffusion pipelines by @fhfuih in https://github.com/vllm-project/vllm-omni/pull/1235
* [Bugfix] Support append/extend/count actions in TrackingArgumentParser by @junhee-yoo in https://github.com/vllm-project/vllm-omni/pull/4639
* [Model] Cosmos3: Skip unused action conditioning frames by @sgillen in https://github.com/vllm-project/vllm-omni/pull/4789
* [image][bugfix] Fix HunyuanImage3 bridge request batching by @Celeste-jq in https://github.com/vllm-project/vllm-omni/pull/4849
* [Bugfix] Fix request batching regression on Ming-flash-omni ImageGen by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4837
* [Diffusion] Add Diffusers quantization config bridge by @CAICAIIs in https://github.com/vllm-project/vllm-omni/pull/3837
* [Feature][Refactor] LTX-2.3 I2V Support, Public Examples, and Shared Pipeline Refactor by @mglyn in https://github.com/vllm-project/vllm-omni/pull/4739
* [BugFix] Use KV cache scale mapper in custom loaders by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4810
* [Bugfix] Fix pytest 9.1.0 duplicate parametrization in run_benchmark.py by @yenuo26 in https://github.com/vllm-project/vllm-omni/pull/4784
* [Bugfix] Report stepwise Qwen-Image profiler metrics by @akshatvishu in https://github.com/vllm-project/vllm-omni/pull/4517
* [ROCm][Perf] Add ROCm support for MXFP4 (w4a4) online quantization by @avjves in https://github.com/vllm-project/vllm-omni/pull/4104
* [NPU] Upgrade to v0.24.0 by @FayeSpica in https://github.com/vllm-project/vllm-omni/pull/4848
* [Feature] Support MOSS-TTS-Local-v1.5 Streaming by @gcanlin in https://github.com/vllm-project/vllm-omni/pull/4804
* [BugFix][Ming-TTS] Improve Seed-TTS WER accuracy by @LHXuuu in https://github.com/vllm-project/vllm-omni/pull/4859
* [BugFix] Adapt to pin_memory in vLLM v0.24.0 by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4860
* [Bugfix] Qwen-Image-Layered: allow layers=2 to match upstream demo by @ktrk115 in https://github.com/vllm-project/vllm-omni/pull/4777
* [BugFix] deploy schema stage runtime overrides for CLI diffusion flags no longer works by @xiaohajiayou in https://github.com/vllm-project/vllm-omni/pull/4101
* [Bugfix] Adapt ZImagePipeline.forward calling in MingImagePipeline with DiffusionRequestBatch  by @yuanheng-zhao in https://github.com/vllm-project/vllm-omni/pull/4866
* [BugFix] Preserve explicit deploy stage engine_extras by @CAICAIIs in https://github.com/vllm-project/vllm-omni/pull/4255
* [Perf][VoxCPM2] Add unified decode graph optimization by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4606
* [Cleanup]Replace legacy logging branches with PreparedRequest.model_type by @sphinxkkkbc in https://github.com/vllm-project/vllm-omni/pull/4862
* [Bugfix][Qwen3-TTS] Seed residual MTP sampling by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4869
* [Bugfix] Fix Qwen3-TTS no-async-chunk payload splitting by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4870
* [Fix][VoxCPM2] Follow up unified graph config gate by @Sy0307 in https://github.com/vllm-project/vllm-omni/pull/4878
* [BDE] Phase 1: engine-level KV cache management for DreamZero (RFC #4366) by @tzhouam in https://github.com/vllm-project/vllm-omni/pull/4534
* [BugFix] Drop removed get_cache_scale call in HunyuanImage3 diffusion loader by @wangrzneu in https://github.com/vllm-project/vllm-omni/pull/4892
* [Bugfix][Qwen3-TTS] Keep seeded residual MTP sampling batched by @linyueqian in https://github.com/vllm-project/vllm-omni/pull/4889
* [Bugfix][sleep mode]: guard generation on partial wake and ensure wake idempotency by @Flink-ddd in https://github.com/vllm-project/vllm-omni/pull/4834
* [Bugfix] Fix HunyuanImage3 MoE groups for vLLM 0.24 by @TaffyOfficial in https://github.com/vllm-project/vllm-omni/pull/4893
* [BugFix] Preserve diffusion extra_body params for out-of-process CFG Parallel by @princepride in https://github.com/vllm-project/vllm-omni/pull/4900
* [BugFix]: Add vae_use_tiling in magi_human test by @princepride in https://github.com/vllm-project/vllm-omni/pull/4877
* [CI][sleep mode] fix sleep pytest CI error by @Flink-ddd in https://github.com/vllm-project/vllm-omni/pull/4912
* [Bugfix] Fix full-payload mm splitting for dual hidden/scheduled batch axes (#4870 follow-up) by @amy-why-3459 in https://github.com/vllm-project/vllm-omni/pull/4910
* [Bugfix] Fix tokenizer config for higgs-audio-v3 test error by @herotai214 in https://github.com/vllm-project/vllm-omni/pull/4918

## New Contributors
* @yangyonggit made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4052
* @nagelanping made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3941
* @n0n4m39911 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4210
* @inkcherry made their first contribution in https://github.com/vllm-project/vllm-omni/pull/1742
* @abinggo made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3774
* @FED4 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4035
* @clodaghwalsh17 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3738
* @kushanam made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4328
* @yuxinyuan made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4346
* @zyz111222 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4283
* @jjuvonen-amd made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3763
* @henryj made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4373
* @yuzhudong made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4282
* @KwokhoTsui made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4452
* @FayeSpica made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4454
* @kTorp made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4293
* @YIHONG-JIN made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4025
* @zeningc made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4026
* @syd520zy made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4490
* @Ronnie-Rui made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3664
* @QianCyrus made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4579
* @MrDongsls made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3862
* @ForestWisdom made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4648
* @dsocek made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3956
* @Acerak01-fy made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4425
* @gglorian made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4192
* @mengker33 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3927
* @rahul-steiger-nv made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4620
* @suyanli220 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4768
* @Nughm3 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4531
* @ayaka14732 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4781
* @moisf56 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4691
* @zzehli made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4607
* @wangrzneu made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4817
* @tthakkal made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4825
* @aporix made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4706
* @junhee-yoo made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4639
* @sgillen made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4789
* @CAICAIIs made their first contribution in https://github.com/vllm-project/vllm-omni/pull/3837
* @ktrk115 made their first contribution in https://github.com/vllm-project/vllm-omni/pull/4777

**Full Changelog**: https://github.com/vllm-project/vllm-omni/compare/v0.22.0...v0.24.0