v0.10.2

## Highlights
This release contains 740 commits from 266 contributors (97 new)!

**Breaking Changes**: This release includes PyTorch 2.8.0 upgrade, V0 deprecations, and API changes - please review the changelog carefully.

**aarch64 support**: This release features native support for aarch64 allowing usage of vLLM on GB200 platform. The docker image `vllm/vllm-openai` should already be multiplatform. To install the wheels, you can download the wheels from this release artifact or install via 
```
uv pip install vllm==0.10.2 --extra-index-url https://wheels.vllm.ai/0.10.2/ --torch-backend=auto
```

### Model Support
* **New model families and enhancements**: Apertus (#23068), LFM2 (#22845), MiDashengLM (#23652), Motif-1-Tiny (#23414), Seed-Oss (#23241), Google EmbeddingGemma-300m (#24318), GTE sequence classification (#23524), Donut OCR model (#23229), KeyeVL-1.5-8B (#23838), R-4B vision model (#23246), Ernie4.5 VL (#22514), MiniCPM-V 4.5 (#23586), Ovis2.5 (#23084), Qwen3-Next with hybrid attention (#24526), InternVL3.5 with video support (#23658), Qwen2Audio embeddings (#23625), NemotronH Nano VLM (#23644), BLOOM V1 engine support (#23488), and Whisper encoder-decoder for V1 (#21088).
* **Pipeline parallelism expansion**: Added PP support for Hunyuan (#24212), Ovis2.5 (#23405), GPT-OSS (#23680), and Kimi-VL-A3B-Thinking-2506 (#23114).
* **Data parallelism for vision models**: Enabled DP for ViT across Qwen2.5VL (#22742), MiniCPM-V (#23948, #23327), Kimi-VL (#23817), and GLM-4.5V (#23168).
* **LoRA ecosystem expansion**: Added LoRA support to Voxtral (#24517), Qwen-2.5-Omni (#24231), and DeepSeek models V2/V3/R1-0528 (#23971), with significantly faster LoRA startup performance (#23777).
* **Classification and pooling enhancements**: Multi-label classification support (#23173), logit bias and sigmoid normalization (#24031), and FP32 precision heads for pooling models (#23810).
* **Performance optimizations**: Removed unnecessary CUDA sync from GLM-4.1V (#24332) and Qwen2VL (#24334) preprocessing, eliminated redundant all-reduce in Qwen3 MoE (#23169), optimized InternVL CPU threading (#24519), and GLM4.5-V video frame decoding (#24161).

### Engine Core
* **V1 engine maturation**: Extended V1 support to compute capability < 8.0 (#23614, #24022), added cross-attention KV cache for encoder-decoder models (#23664), request-level logits processor integration (#23656), and KV events from connectors (#19737).
* **Backend expansion**: Terratorch backend integration (#23513) enabling non-language model tasks like semantic segmentation and geospatial applications with `--model-impl terratorch` support.
* **Hybrid and Mamba model improvements**: Enabled full CUDA graphs by default for hybrid models (#22594), disabled prefix caching for hybrid/Mamba models (#23716), added FP32 SSM kernel support (#23506), full CUDA graph support for Mamba1 (#23035), and V1 as default for Mamba models (#23650).
* **Performance core improvements**: `--safetensors-load-strategy` for NFS based file loading acceleration (#24469), critical CUDA graph capture throughput fix (#24128), scheduler optimization for single completions (#21917), multi-threaded model weight loading (#23928), and tensor core usage enforcement for FlashInfer decode (#23214).
* **Multimodal enhancements**: Multimodal cache tracking with mm_hash (#22711), UUID-based multimodal identifiers (#23394), improved V1 video embedding estimation (#24312), and simplified multimodal UUID handling (#24271).
* **Sampling and structured outputs**: Support for all prompt logprobs (#23868), final logprobs (#22387), grammar bitmask optimization (#23361), and user-configurable KV cache memory size (#21489).
* **Distributed**: Support Decode Context Parallel (DCP) for MLA (#23734)

### Hardware & Performance
* **NVIDIA Blackwell/SM100 generation**: FP8 MLA support with CUTLASS backend (#23289), DeepGEMM Linear with 1.5% E2E throughput improvement (#23351), Hopper DeepGEMM E8M0 for DeepSeekV3.1 (#23666), SM100 FlashInfer CUTLASS MoE FP8 backend (#22357), MXFP4 fused CUTLASS MoE (#23696), default MXFP4 MoE on Blackwell (#23008), and GPT-OSS DP/EP support with 52,003 tokens/s throughput (#23608).
* **Breaking change**: FlashMLA disabled on Blackwell GPUs due to compatibility issues (#24521).
* **Kernel and attention optimizations**: FlashAttention MLA with CUDA graph support (#14258, #23958), V1 cross-attention support (#23297), FP8 support for FlashMLA (#22668), fused grouped TopK for MoE (#23274), Flash Linear Attention kernels (#24518), and W4A8 support on Hopper (#23198).
* **Performance improvements**: 13.7x speedup for token conversion (#20413), TTIT/TTFT improvements for disaggregated serving (#22760), symmetric memory all-reduce by default (#24111), FlashInfer warmup during startup (#23439), V1 model execution overlap (#23569), and various Triton configuration tuning (#23748, #23939).
* **Platform expansion**: Apple Silicon bfloat16 support for M2+ (#24129), IBM Z V1 engine support (#22725), Intel XPU torch.compile (#22609), XPU MoE data parallelism (#22887), XPU Triton attention (#24149), XPU FP8 quantization (#23148), and ROCm pipeline parallelism with Ray (#24275).
* **Model-specific optimizations**: Hardware-tuned MoE configurations for Qwen3-Next on B200/H200/H100 (#24698, #24688, #24699, #24695), GLM-4.5-Air-FP8 B200 configs (#23695), Kimi K2 optimization (#24597), and QWEN3 Coder/Thinking configs (#24266, #24330).

### Quantization
* **New quantization capabilities**: Per-layer quantization routing (#23556), GGUF quantization with layer skipping (#23188), NFP4+FP8 MoE support (#22674), W4A8 channel scales (#23570), and AMD CDNA2/CDNA3 FP4 support (#22527).
* **Advanced quantization infrastructure**: Compressed tensors transforms for linear operations (#22486) enabling techniques like SpinQuantR1R2R4 and QuIP quantization methods.
* **FlashInfer quantization integration**: FP8 KV cache for TRTLLM prefill attention (#24197), FP8-qkv attention kernels (#23647), and FP8 per-tensor GEMMs (#22895).
* **Platform-specific quantization**: ROCm TorchAO quantization enablement (#24400) and TorchAO module swap configuration (#21982).
* **Performance optimizations**: MXFP4 MoE loading cache optimization (#24154) and compressed tensors version updates (#23202).
* **Breaking change**: Removed original Marlin quantization format (#23204).

### API & Frontend
* **OpenAI API enhancements**: Gemma3n audio transcription/translation endpoints (#23735), transcription response usage statistics (#23576), and return_token_ids parameter (#22587).
* **Response API improvements**: Streaming support for non-harmony responses (#23741), non-streaming logprobs (#23319), MCP tool background mode (#23494), MCP streaming+background support (#23927), and tool output token reporting (#24285).
* **Frontend optimizations**: Error stack traces with --log-error-stack (#22960), collective RPC endpoint (#23075), beam search concurrency optimization (#23599), unnecessary detokenization skipping (#24236), and custom media UUIDs (#23449).
* **Configuration enhancements**: Formalized --mm-encoder-tp-mode flag (#23190), VLLM_DISABLE_PAD_FOR_CUDAGRAPH environment variable (#23595), EPLB configuration parameter (#20562), embedding endpoint chat request support (#23931), and LM Format Enforcer V1 integration (#22564).

### Dependencies
* **Major updates**: PyTorch 2.8.0 upgrade (#20358) - breaking change requiring environment updates, FlashInfer v0.3.0 upgrade (#24086), and FlashInfer 0.2.14.post1 maintenance update (#23537).
* **Supporting updates**: XGrammar 0.1.23 (#22988), TPU core dump fix with tpu_info 0.4.0 (#23135), and compressed tensors version bump (#23202).
* **Deployment improvements**: FlashInfer cubin directory environment variable (#22675) for offline environments and pre-cached CUDA binaries.

### V0 Deprecation
* **Backend removals**: V0 Neuron backend deprecation (#21159), V0 pooling model support removal (#23434), V0 FlashInfer attention backend removal (#22776), and V0 test cleanup (#23418, #23862).
* **API breaking changes**: prompt_token_ids fallback removal from LLM.generate and LLM.embed (#18800), LoRA extra vocab size deprecation warning (#23635), LoRA bias parameter deprecation (#24339), and metrics naming change from TPOT to ITL (#24110).

### Breaking Changes
1. **PyTorch 2.8.0 upgrade** - Environment dependency change requiring updated CUDA versions
2. **FlashMLA Blackwell restriction** - FlashMLA disabled on Blackwell GPUs due to compatibility issues
3. **V0 feature removals** - Neuron backend, pooling models, FlashInfer attention backend
5. **Quantizations** - Removed quantized Mixtral hack implementation, and original Marlin format. 
6. **Metrics renaming** - TPOT deprecated in favor of ITL

## What's Changed
* [Misc] Minor code cleanup for _get_prompt_logprobs_dict by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23064
* [Misc] enhance static type hint by @andyxning in https://github.com/vllm-project/vllm/pull/23059
* [Bugfix] fix Qwen2.5-Omni processor output mapping by @DoubleVII in https://github.com/vllm-project/vllm/pull/23058
* [Bugfix][CI] Machete kernels: deterministic ordering for more cache hits by @andylolu2 in https://github.com/vllm-project/vllm/pull/23055
* [Misc] refactor function name by @andyxning in https://github.com/vllm-project/vllm/pull/23029
* [Misc] Fix backward compatibility from #23030 by @ywang96 in https://github.com/vllm-project/vllm/pull/23070
* [XPU] Fix compile size for xpu by @jikunshang in https://github.com/vllm-project/vllm/pull/23069
* [XPU][CI]add xpu env vars in CI scripts by @jikunshang in https://github.com/vllm-project/vllm/pull/22946
* [Refactor] Define MultiModalKwargsItems separate from MultiModalKwargs by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23053
* [Bugfix] fix IntermediateTensors equal method by @andyxning in https://github.com/vllm-project/vllm/pull/23027
* [Refactor] Get prompt updates earlier by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23097
* chore: remove unnecessary patch_padding_side for the chatglm model by @carlory in https://github.com/vllm-project/vllm/pull/23090
* [Bugfix] Support compile for Transformers multimodal by @zucchini-nlp in https://github.com/vllm-project/vllm/pull/23095
* [CI Bugfix] Pin `openai<1.100` to unblock CI by @mgoin in https://github.com/vllm-project/vllm/pull/23118
* fix: OpenAI SDK compat (ResponseTextConfig) by @h-brenoskuk in https://github.com/vllm-project/vllm/pull/23126
* Use Blackwell FlashInfer MXFP4 MoE by default if available  by @mgoin in https://github.com/vllm-project/vllm/pull/23008
* Install tpu_info==0.4.0 to fix core dump for TPU by @xiangxu-google in https://github.com/vllm-project/vllm/pull/23135
* [Misc] Minor refactoring for prepare_inputs by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23116
* [Spec Decode] Make `propose_draft_token_ids` non-blocking for lower TTFT by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23041
* [Misc] Add @tdoublep as a maintainer of hybrid model and Triton-attention related code by @tdoublep in https://github.com/vllm-project/vllm/pull/23122
* [CI][V0 Deprecation] Removed V0 Only Chunked Prefill and Prefix Caching Tests by @robertgshaw2-redhat in https://github.com/vllm-project/vllm/pull/22871
* [V0 Deprecation] Remove V0 FlashInfer attention backend by @WoosukKwon in https://github.com/vllm-project/vllm/pull/22776
* chore: disable enable_cpp_symbolic_shape_guards by @xiszishu in https://github.com/vllm-project/vllm/pull/23048
* [TPU] make ptxla not imported when using tpu_commons by @yaochengji in https://github.com/vllm-project/vllm/pull/23081
* [Hardware][IBM Z]Enable v1 for s390x and s390x dockerfile fixes by @nikheal2 in https://github.com/vllm-project/vllm/pull/22725
* Migrate InternVLImagePixelInputs (in nemotron_vl.py) to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/22023
* [Log] Warning Once for Cutlass MLA  by @yewentao256 in https://github.com/vllm-project/vllm/pull/23137
* [Model] Support Pipeline Parallelism for moonshotai/Kimi-VL-A3B-Thinking-2506 by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23114
* [misc] split engine_model into json file for nsys profile tool by @gracehonv in https://github.com/vllm-project/vllm/pull/23117
* [Benchmark] Add flag --served-model-name to benchmark_serving_multi_turn by @pliops-daniels in https://github.com/vllm-project/vllm/pull/22889
* Fix GLM-4.5V-FP8 numerical issue by @zixi-qi in https://github.com/vllm-project/vllm/pull/22949
* [Misc] Add request_id into benchmark_serve.py by @hustxiayang in https://github.com/vllm-project/vllm/pull/23065
* [Bugfix] Fix broken Minimax-01-VL model by @Isotr0py in https://github.com/vllm-project/vllm/pull/22116
* [bug fix] Fix llama4 spec decoding by @zixi-qi in https://github.com/vllm-project/vllm/pull/22691
* [Misc] Avoid accessing req_ids inside a loop by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23159
* [Doc] use power of 2 by @Tialo in https://github.com/vllm-project/vllm/pull/23172
* [Misc] Fix seq_lens for graph capture by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23175
* [NVIDIA] Support Flashinfer TRTLLM FP8-q/kv/out Attention Kernel by @elvischenv in https://github.com/vllm-project/vllm/pull/21716
* [Model] Add transformers problem_type (e.g. multi_label_classification) support by @noooop in https://github.com/vllm-project/vllm/pull/23173
* [Model] support new model ovis2.5 by @myselvess in https://github.com/vllm-project/vllm/pull/23084
* [Bugfix] Fix benchmark_moe.py  by @jeejeelee in https://github.com/vllm-project/vllm/pull/23177
* [FEAT] [Performance] Enable DP for ViT in Qwen2.5VL by @tjtanaa in https://github.com/vllm-project/vllm/pull/22742
* [Model] Removes redundant all-reduce operation in Qwen3MoeSparseMoeBlock by @yiz-liu in https://github.com/vllm-project/vllm/pull/23169
* Add return_token_ids parameter to OpenAI API endpoints by @ultmaster in https://github.com/vllm-project/vllm/pull/22587
* Migrate LlavaOnevisionMultiInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/21844
* [CI/Build] Update transformers to v4.55.2 by @Isotr0py in https://github.com/vllm-project/vllm/pull/23093
* [Misc] Fix the benchmark's README and improve the error messages for the benchmark's argument checks by @tanruixiang in https://github.com/vllm-project/vllm/pull/22654
* [Frontend] Add `/collective_rpc` API endpoint by @22quinn in https://github.com/vllm-project/vllm/pull/23075
* [Misc] Enable yapf for FlashInfer backend by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23193
* [Bugfix] Fix accuracy issue when using flashinfer cutlass moe, TP=1 and modelopt. by @bnellnm in https://github.com/vllm-project/vllm/pull/23125
* fix: use cache_salt for gpt-oss by @dr75 in https://github.com/vllm-project/vllm/pull/23186
* [Misc] Minor refactoring for FlashInfer backend by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23147
* [CI/Build] Add support for Python 3.13 by @mgoin in https://github.com/vllm-project/vllm/pull/13164
* [NVIDIA] Add SM100 Flashinfer Cutlass MoE fp8 backend by @amirkl94 in https://github.com/vllm-project/vllm/pull/22357
* [CI/Build] Replace lm-eval gsm8k tests with faster implementation by @mgoin in https://github.com/vllm-project/vllm/pull/23002
* [BugFix] fix CUTLASS MLA full cudagraph  by @LucasWilkinson in https://github.com/vllm-project/vllm/pull/23200
* [Benchmarks] Add video inputs to ShareGPTDataset.  by @huachenheli in https://github.com/vllm-project/vllm/pull/23199
* [Quantization] Bump Compressed Tensors Version by @kylesayrs in https://github.com/vllm-project/vllm/pull/23202
* [Core] Optimize scheduler request removal for single completions by @chi2liu in https://github.com/vllm-project/vllm/pull/21917
* [CI Perf] Only test bfloat16 for tests/compile/test_fusion_all_reduce.py by @mgoin in https://github.com/vllm-project/vllm/pull/23132
* [Core] Add torch profiler CPU traces for AsyncLLM. by @huachenheli in https://github.com/vllm-project/vllm/pull/21794
* [Doc] Update V1 status of various pooling models by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23189
* [Attention] Optimize make_local_attention_virtual_batches for Flash Attention by @linzebing in https://github.com/vllm-project/vllm/pull/23185
* Fix a performance comparison issue in Benchmark Suite by @louie-tsai in https://github.com/vllm-project/vllm/pull/23047
* chore: support pytorch format in lora  by @KilJaeeun in https://github.com/vllm-project/vllm/pull/22790
* [CI/Build] Also check DP in benchmarks throughput script by @zhewenl in https://github.com/vllm-project/vllm/pull/23038
* [CI/Build] Sync multimodal tests by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23181
* [BugFix] Fix stuck stats/metrics after requests are aborted by @njhill in https://github.com/vllm-project/vllm/pull/22995
* fix cuda graph by @fsx950223 in https://github.com/vllm-project/vllm/pull/22721
* [Model] use autoWeightsLoader for gptoss by @calvin0327 in https://github.com/vllm-project/vllm/pull/22446
* Fix missing quotes by @wzshiming in https://github.com/vllm-project/vllm/pull/23242
* [Model] Support deepseek with eagle by @xyang16 in https://github.com/vllm-project/vllm/pull/21086
* [Bugfix] Ensure correctness of Cohere2Vision processing by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23245
* Update to flashinfer-python==0.2.12 and disable AOT compile for non-release image by @mgoin in https://github.com/vllm-project/vllm/pull/23129
* [Model][V1] Support Ernie MTP by @xyxinyang in https://github.com/vllm-project/vllm/pull/22169
* [Model] Improve olmo and olmo2 by @jeejeelee in https://github.com/vllm-project/vllm/pull/23228
* [Fix] fix offline env use local mode path by @lengrongfu in https://github.com/vllm-project/vllm/pull/22526
* [Bugfix] Ensure correctness of HCXVision processing by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23254
* [Kernel] CUTLASS MoE FP8: Integrate cuda moe permute/unpermute by @shixianc in https://github.com/vllm-project/vllm/pull/23045
* [CLI][Doc] Formalize `--mm-encoder-tp-mode` by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23190
* [Misc] Add max_seq_len to CommonAttentionMetadata  by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23216
* [FIXBUG ] Allow disabling rocm_aiter_fa backend for ROCm GPUs not compatible with AITER by @JartX in https://github.com/vllm-project/vllm/pull/22795
* Support conditional torch.compile per module by @sarckk in https://github.com/vllm-project/vllm/pull/22269
* Migrate Mistral3ImagePixelInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/21945
* Limit HTTP header count and size by @russellb in https://github.com/vllm-project/vllm/pull/23267
* Small fix for Command-A-Vision by @dongluw in https://github.com/vllm-project/vllm/pull/23268
* [Kernel/Quant] Remove the original marlin format and qqq by @mgoin in https://github.com/vllm-project/vllm/pull/23204
* [Fix] correct tool_id for kimi-k2 when use tool_choice=required by @MoyanZitto in https://github.com/vllm-project/vllm/pull/21259
* [Frontend] improve error logging of chat completion by @heheda12345 in https://github.com/vllm-project/vllm/pull/22957
* [Optimization] Speed up function `_convert_tokens_to_string_with_added_encoders` by 13.7x by @misrasaurabh1 in https://github.com/vllm-project/vllm/pull/20413
* Do not use eval() to convert unknown types by @russellb in https://github.com/vllm-project/vllm/pull/23266
* [Feature] use --eplb_config to set eplb param by @lengrongfu in https://github.com/vllm-project/vllm/pull/20562
* [misc] fix multiple arch wheels for the nightly index by @youkaichao in https://github.com/vllm-project/vllm/pull/23110
* Remove chunked_prefill_enabled flag in V1 MLA by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/23183
* Feature/mla tests by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/23195
* [Fix] remove is_marlin param in benchmark_moe by @shixianc in https://github.com/vllm-project/vllm/pull/23286
* [EP] Add logging for experts map by @22quinn in https://github.com/vllm-project/vllm/pull/22685
* Remove duplicate entry in vllm.attention.__all__ by @russellb in https://github.com/vllm-project/vllm/pull/23296
* [CI Bugfix] Fix CI by fully removing --enable-prompt-adapter by @mgoin in https://github.com/vllm-project/vllm/pull/23284
* [Optimization] Make new_block_ids None if empty by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23262
* [CPU] Refactor CPU W8A8 scaled_mm by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/23071
* [CI/Build] Split out mm processor tests by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23260
* [V1][Mamba1] - Full CUDA and Piecewise CUDA Graphs Support by @Josephasafg in https://github.com/vllm-project/vllm/pull/23035
* [Compile] Fix Compile Warning SM100 Cutlass MLA by @yewentao256 in https://github.com/vllm-project/vllm/pull/23287
* [Model][VLM] Support R-4B Model by @yannqi in https://github.com/vllm-project/vllm/pull/23246
* Delete images older than 24h. by @QiliangCui in https://github.com/vllm-project/vllm/pull/23291
* [CI] Block the cu126 wheel build while broken by @mgoin in https://github.com/vllm-project/vllm/pull/23285
* [Sampler] Support returning final logprobs by @22quinn in https://github.com/vllm-project/vllm/pull/22387
* [Bugfix] Fix extra whitespace in strings caused by newline by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23272
* [BugFix] Fix Python 3.9 Support by @jaredoconnell in https://github.com/vllm-project/vllm/pull/23306
* [Model] Add LFM2 architecture by @paulpak58 in https://github.com/vllm-project/vllm/pull/22845
* [Refactor] Simplify code for MM budget by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23310
* [Doc] Fix batch-level DP example by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23325
* [Performance] V1 Pooling Models E2E Performance Optimization by @noooop in https://github.com/vllm-project/vllm/pull/23162
* [V1] Remove unnecessary check for main thread by @robertgshaw2-redhat in https://github.com/vllm-project/vllm/pull/23298
* [Bugfix] set system_message in phi4mini chat template by @zhuangqh in https://github.com/vllm-project/vllm/pull/23309
* [Multimodal] Always enable hashing mm data by @ywang96 in https://github.com/vllm-project/vllm/pull/23308
* [ci/build] Fix abi tag for aarch64 by @youkaichao in https://github.com/vllm-project/vllm/pull/23329
* Migrate MolmoImageInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/22022
* Fix nvfp4 swizzling by @yiliu30 in https://github.com/vllm-project/vllm/pull/23140
* add tg-mxfp4-moe-test by @IwakuraRein in https://github.com/vllm-project/vllm/pull/22540
* [Bug] Fix R1 Accuracy 0 Bug by @yewentao256 in https://github.com/vllm-project/vllm/pull/23294
* [Bugfix] Fix port conflict by obtaining a list of open ports upfront by @minosfuture in https://github.com/vllm-project/vllm/pull/21894
* [Misc] Misc code cleanup/simplification by @njhill in https://github.com/vllm-project/vllm/pull/23304
* [BugFix][gpt-oss] Fix Chat Completion with Multiple Output Message by @heheda12345 in https://github.com/vllm-project/vllm/pull/23318
* [Misc] fix VLLM_TORCH_PROFILER_DIR to absolute path by @andyxning in https://github.com/vllm-project/vllm/pull/23191
* [Core] Always use tensor cores for Flashinfer Decode Wrapper by @pavanimajety in https://github.com/vllm-project/vllm/pull/23214
* Make sure that vectorize_with_alignment produced vectorized global loads by @elvircrn in https://github.com/vllm-project/vllm/pull/23182
* [Structured Outputs] Refactor bitmask construction into get_grammar_bitmask by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23361
* [CI] Clean up actions: remove helm, publish workflows and improve pr … by @simon-mo in https://github.com/vllm-project/vllm/pull/23377
* [CI] improve pr comments bot by @simon-mo in https://github.com/vllm-project/vllm/pull/23380
* [Perf] Small optimizations for silu_mul_fp8_quant_deep_gemm by @mgoin in https://github.com/vllm-project/vllm/pull/23265
* Always use cache mounts when installing vllm to avoid populating pip cache in the image. Also remove apt cache. by @tvalentyn in https://github.com/vllm-project/vllm/pull/23270
* [Feature][Responses API] Support logprobs(non-stream) by @kebe7jun in https://github.com/vllm-project/vllm/pull/23319
* [Core] Support custom executor qualname by @22quinn in https://github.com/vllm-project/vllm/pull/23314
* [Kernel] Add FP8 support with FlashMLA backend by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/22668
* [Deprecation] Remove `prompt_token_ids` arg fallback in `LLM.generate` and `LLM.embed` by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/18800
* Migrate MllamaImagePixelInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/22020
* [CI/Build] Skip Idefics3 and SmolVLM generation test again by @Isotr0py in https://github.com/vllm-project/vllm/pull/23356
* [Feature] Enable DeepGEMM Linear on B200; 1.5% E2E throughput improvement by @yewentao256 in https://github.com/vllm-project/vllm/pull/23351
* [CI] Add end-to-end V1 min_tokens test coverage by @arjunbreddy22 in https://github.com/vllm-project/vllm/pull/22495
* [Misc] Add gemma3 chat template with pythonic-style function calling by @philipchung in https://github.com/vllm-project/vllm/pull/17149
* [New Model] Add Seed-Oss model by @FoolPlayer in https://github.com/vllm-project/vllm/pull/23241
* [Attention] Refactor AttentionMetadata Preparation for Encoder-only Models by @heheda12345 in https://github.com/vllm-project/vllm/pull/23154
* [P/D][Nixl] Make kv cache register compatible with hybrid memory allocator by @sfeng33 in https://github.com/vllm-project/vllm/pull/23079
* [gpt-oss] add input/output usage in responses api when harmony context is leveraged by @gcalmettes in https://github.com/vllm-project/vllm/pull/22667
* Migrate MiniCPMOAudioInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/21847
* [Bugfix] Fix pooling models on non-CUDA devices by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/23392
* [V0 Deprecation] Remove V0 LoRA test by @jeejeelee in https://github.com/vllm-project/vllm/pull/23418
* [Misc] Move M-RoPE init logic to _init_mrope_positions by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23422
* [Attention] Allow V1 flash_attn to support cross-attention by @russellb in https://github.com/vllm-project/vllm/pull/23297
* [misc] Remove outdate comment about runai_model_streamer by @carlory in https://github.com/vllm-project/vllm/pull/23421
* [Doc] Update the doc for log probs + prefix caching by @heheda12345 in https://github.com/vllm-project/vllm/pull/23399
* [Misc] local import code clean by @andyxning in https://github.com/vllm-project/vllm/pull/23420
* [Bug fix] Dynamically setting the backend variable for genai_perf_tests in the run-nightly-benchmark script by @namanlalitnyu in https://github.com/vllm-project/vllm/pull/23375
* [Fix] Bump triton version in rocm-build requirements by @bringlein in https://github.com/vllm-project/vllm/pull/21630
* [Bugfix]: Installing dev environment due to pydantic incompatible version by @hickeyma in https://github.com/vllm-project/vllm/pull/23353
* [Speculators][Speculative Decoding] Fix Qwen 2 Eagle3 Support by @PapaGoose in https://github.com/vllm-project/vllm/pull/23337
* [BugFix] Fix the issue where image embeddings were incorrectly split.… by @bppps in https://github.com/vllm-project/vllm/pull/23366
* fix(tests): Ensure reliable CUDA cache clearing in MoE test by @AzizCode92 in https://github.com/vllm-project/vllm/pull/23416
* Add unit tests for batched guided and non-guided requests by @sarckk in https://github.com/vllm-project/vllm/pull/23389
* [Doc]: fix various typos in multiple files by @didier-durand in https://github.com/vllm-project/vllm/pull/23179
* [Model] Add Ovis2.5 PP support by @Isotr0py in https://github.com/vllm-project/vllm/pull/23405
* [Bugfix] Fix broken Florence-2 model by @Isotr0py in https://github.com/vllm-project/vllm/pull/23426
* [Quantization] Allow GGUF quantization to skip unquantized layer by @Isotr0py in https://github.com/vllm-project/vllm/pull/23188
* add an env var for path to pre-downloaded flashinfer cubin files by @842974287 in https://github.com/vllm-project/vllm/pull/22675
* [CI/Build] add EP dependencies to docker by @zhewenl in https://github.com/vllm-project/vllm/pull/21976
* [PERF] PyTorch Symmetric Memory All-Reduce by @ilmarkov in https://github.com/vllm-project/vllm/pull/20759
* [BugFix][AMD][Quantization] Fix torch.compile issue where wvSplitKQ not being called when it should when using quantized FP8 model by @rasmith in https://github.com/vllm-project/vllm/pull/22281
* [NVIDIA] Support Flashinfer TRTLLM FP8-q/kv NVFP4-out Attention Kernel by @elvischenv in https://github.com/vllm-project/vllm/pull/22703
* [BugFix] Fix batch updates for pooling models by @njhill in https://github.com/vllm-project/vllm/pull/23398
* [BugFix] Fix `MinPLogitsProcessor.update_states()` by @njhill in https://github.com/vllm-project/vllm/pull/23401
* [Model] Support DP for ViT on MiniCPM-V-4 by @david6666666 in https://github.com/vllm-project/vllm/pull/23327
* [UX] Move Dockerfile DeepGEMM install to tools/install_deepgemm.sh by @mgoin in https://github.com/vllm-project/vllm/pull/23360
* Quantization: support FP4 quantized models on AMD CDNA2/CDNA3 GPUs by @fengli1702 in https://github.com/vllm-project/vllm/pull/22527
* Add glm4.5v tp2,4 fp8 config on H100_80GB by @chenxi-yang in https://github.com/vllm-project/vllm/pull/23443
* Revert "[PERF] Use faster way of decode in tokenizer: avoid useless list-to-list conversion (#20000)" by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23396
* fix(tests): Correct unreachable assertion in truncation test by @AzizCode92 in https://github.com/vllm-project/vllm/pull/23425
* Support DeepSeek-V3.1 tool call by @Xu-Wenqing in https://github.com/vllm-project/vllm/pull/23454
* [Misc] Modify CacheConfig import by @jeejeelee in https://github.com/vllm-project/vllm/pull/23459
* [gpt-oss] Streaming Output for Python Tool by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23409
* Migrate Pixtral inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23472
* [Bugfix] Add strong reference to CUDA pluggable allocator callbacks by @22quinn in https://github.com/vllm-project/vllm/pull/23477
* Migrate Paligemma inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23470
* [kernel] Support W4A8 on Hopper by @czhu-cohere in https://github.com/vllm-project/vllm/pull/23198
* [Misc] update dict parse to EPLBConfig from json dumps to dict unpacking by @lengrongfu in https://github.com/vllm-project/vllm/pull/23305
* (Misc): add missing test for zero truncation size. by @teekenl in https://github.com/vllm-project/vllm/pull/23457
* [New Model]Donut model by @princepride in https://github.com/vllm-project/vllm/pull/23229
* [Model] Enable BLOOM on V1 by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23488
* [Misc] Remove unused slot_mapping buffer by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23502
* fix incompatibililty with non cuda platform for nvfp4 by @luccafong in https://github.com/vllm-project/vllm/pull/23478
* [Doc: ]fix various typos in multiple files by @didier-durand in https://github.com/vllm-project/vllm/pull/23487
* [Perf] Add Triton config for DeepSeek V3 FP8 EP32 H200 by @minosfuture in https://github.com/vllm-project/vllm/pull/23504
* Frontend: Adding LM Format Enforcer support to V1 engine by @noamgat in https://github.com/vllm-project/vllm/pull/22564
* [Bugfix] Fix Qwen2.5-VL quantized model weights loading by @zifeitong in https://github.com/vllm-project/vllm/pull/23512
* [Misc] Unified linear print info by @jeejeelee in https://github.com/vllm-project/vllm/pull/23516
* Migrate tarsier inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23500
* Migrate skyworkr1v inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23499
* Migrate DonutImagePixelInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23509
* [Bugfix] Fix Dense module loading for sentence-transformers embedding models (simplified V2) by @FFFfff1FFFfff in https://github.com/vllm-project/vllm/pull/23408
* [gpt-oss] use reasoning channel for reasoning text in serving_chat by @yuguo68 in https://github.com/vllm-project/vllm/pull/22920
* [Refactor] Dynamic `target` and `content` for prompt updates by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23411
* [Core][Multimodal] Track encode cache entries by mm_hash and enable embedding sharing between requests by @fake0fan in https://github.com/vllm-project/vllm/pull/22711
* [Fix] DeepSeek V3.1 tool parser error message by @skyloevil in https://github.com/vllm-project/vllm/pull/23492
* Feature/benchmark/random mm data/images by @h-brenoskuk in https://github.com/vllm-project/vllm/pull/23119
* [Bugfix] Allow dynamic number of patches for llava_onevision by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23525
* [misc] add shanghai meetup by @youkaichao in https://github.com/vllm-project/vllm/pull/23535
* [Attention] Unify mamba and attention backend selection by @ayushsatyam146 in https://github.com/vllm-project/vllm/pull/23171
* [Doc] Add caution for API server scale-out by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23550
* [Refactor] Pass `tokenizer` explicitly instead of binding to prompt update by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23542
* Updates to Flex + VLLm integration by @drisspg in https://github.com/vllm-project/vllm/pull/21416
* [Bugfix] Fix Qwen3 MoE GPTQ inference by @Isotr0py in https://github.com/vllm-project/vllm/pull/23490
* [Refactor] Refactor persistent buffers with CpuGpuBuffer  by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23515
* [test][RL] Add sleep level 2 test and fix reload with sleep mode by @22quinn in https://github.com/vllm-project/vllm/pull/23521
* [Kernel] Add fused grouped_topk kernel for MoE by @xyang16 in https://github.com/vllm-project/vllm/pull/23274
* [Bugfix][V1][P/D]Fix the issue where repeated requests for the same input produce abnormal outputs for P2pNcclConnector by @Abatom in https://github.com/vllm-project/vllm/pull/23403
* [XPU] Delay BF16 check to worker init for spawn compatibility by @chaojun-zhang in https://github.com/vllm-project/vllm/pull/22979
* [TPU][Bugfix] Fixes prompt_token_ids error in tpu tests. by @patemotter in https://github.com/vllm-project/vllm/pull/23574
* [Docs] Update Documentation of Cohere Command-A Models by @Terrencezzj in https://github.com/vllm-project/vllm/pull/23584
* [Misc] Simplify FlashInfer attention metadata by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23585
* [Misc] Add release note draft to PR template by @simon-mo in https://github.com/vllm-project/vllm/pull/23598
* [CI Fix] Pin deepep and pplx tags in tools/ep_kernels/, gate multigpu tests by @mgoin in https://github.com/vllm-project/vllm/pull/23568
* Update Flashinfer to  0.2.14.post1 by @weireweire in https://github.com/vllm-project/vllm/pull/23537
* [Bug] Fix DeepGEMM Env Control by @yewentao256 in https://github.com/vllm-project/vllm/pull/23591
* [CI/Build] Use vLLM client's user agent to fetch images by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23561
* Remove graph_pool as member of VllmBackend and argument to CUDAGraphWrapper by @Copilot in https://github.com/vllm-project/vllm/pull/23385
* [Disagg][Perf] Use CUDA event sync instead of blocking `tolist` to avoid unintentional copy ops blocking across different CUDA streams, improving disagg TTIT/TTFT by @liuzijing2014 in https://github.com/vllm-project/vllm/pull/22760
* [CI/Build] Fix typo in #23561 by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23616
* [fix] fix seed-oss-parser by @FoolPlayer in https://github.com/vllm-project/vllm/pull/23560
* [mypy] Fix incorrect type hint for EAGLE3 support by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23617
* [Benchmarks] add benchmark for embedding models by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23000
* [Docs] Fix titles for multi-file examples that are rendered in the docs by @hmellor in https://github.com/vllm-project/vllm/pull/23573
* Fix CLI parameter documentation inconsistency in pooling_models.md by @oneraghavan in https://github.com/vllm-project/vllm/pull/23630
* [Bugfix] Fix Qwen25VL packed_modules_mapping by @jeejeelee in https://github.com/vllm-project/vllm/pull/23604
* [Bugfix] Fix scheduling when repeated images in one request by @ywang96 in https://github.com/vllm-project/vllm/pull/23544
* [V1] Enable V1 for compute capability < 8.0 + FP32 by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23614
* Fix nits from #20059 by @hmellor in https://github.com/vllm-project/vllm/pull/23548
* Fix writing benchmark results with tuple keys by @huydhn in https://github.com/vllm-project/vllm/pull/23633
* [Perf] Remove duplicated NVFP4 blockscales to save memory by @mgoin in https://github.com/vllm-project/vllm/pull/23379
* [Model] fix DeepSeek e_score_correction_bias dtype to fp32 by @jeejeelee in https://github.com/vllm-project/vllm/pull/23640
* [Bugfix] Add missing enable_log_outputs parameter to init_app_state function by @lordmathis in https://github.com/vllm-project/vllm/pull/23634
* feat: add usage to TranscriptionResponse (text and json response_format) by @gcalmettes in https://github.com/vllm-project/vllm/pull/23576
* Support FlashAttention Backend for Hybrid SSM Models by @heheda12345 in https://github.com/vllm-project/vllm/pull/23299
* [Docs] Fix broken links to `docs/api/summary.md` by @hmellor in https://github.com/vllm-project/vllm/pull/23637
* [Hardware][Mac] Fix the installation fail for Apple Silicon (CPU)  by @OYE93 in https://github.com/vllm-project/vllm/pull/23565
* [Kernel] Added flashinfer fp8 per-tensor gemms by @nvjullin in https://github.com/vllm-project/vllm/pull/22895
* [Doc]: fix various spelling issues in multiple files by @didier-durand in https://github.com/vllm-project/vllm/pull/23636
* [CPU] add cpu fused moe pytorch native implementation by @TianyuLi0 in https://github.com/vllm-project/vllm/pull/23146
* [ROCm] Starting to add AMD code reviewers for ROCm components by @hongxiayang in https://github.com/vllm-project/vllm/pull/23496
* [Docs] Reduce requirements for docs build by @hmellor in https://github.com/vllm-project/vllm/pull/23651
* [Bugfix] fix bf16 multimodal model hash by @yuekaizhang in https://github.com/vllm-project/vllm/pull/23623
* [model] support qwen2audio embedding input by @yuekaizhang in https://github.com/vllm-project/vllm/pull/23625
* [Misc] Add override for allreduce fusion thresholds by @nvjullin in https://github.com/vllm-project/vllm/pull/23639
* [CI] [Doc]: Add GH Action for auto labeling issues with `rocm` tag by @vllmellm in https://github.com/vllm-project/vllm/pull/20988
* [Bugfix] Fix cuda event usage with CPU model runner by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/23643
* [Docs] Fix warnings in `mkdocs build` by @Zerohertz in https://github.com/vllm-project/vllm/pull/23649
* [Docs] [V1] [Hybrid] Update docs to remove FlashInfer constraint for hybrid models by @tdoublep in https://github.com/vllm-project/vllm/pull/23665
* [v1] Add cross-attention KV cache support for encoder-decoder models by @russellb in https://github.com/vllm-project/vllm/pull/23664
* [Bugfix] Fix incorrect original shape in hashing by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23672
* [Misc] Fix comments in `tests/kernels/quantization` by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23675
* [Model] Enable video support for InternVL3.5 models by @Isotr0py in https://github.com/vllm-project/vllm/pull/23658
* [doc] Hybrid KV Cache Manager design doc by @heheda12345 in https://github.com/vllm-project/vllm/pull/22688
* Enhance the pre-notification policy by @sidhpurwala-huzaifa in https://github.com/vllm-project/vllm/pull/23532
* [Docs] Move quant supported hardware table to README by @hmellor in https://github.com/vllm-project/vllm/pull/23663
* [V1][P/D]P2pNcclConnector supports flashinfer by @Abatom in https://github.com/vllm-project/vllm/pull/23536
* [V1] [Hybrid] Enable Full CUDA graph by default for hybrid models in V1 by @tdoublep in https://github.com/vllm-project/vllm/pull/22594
* [Compile] Fix Cmake Warning by @yewentao256 in https://github.com/vllm-project/vllm/pull/23689
* [Bugfix] UnboundLocalError when GptOss reasoning specified by @coval3nte in https://github.com/vllm-project/vllm/pull/23054
* feat: add triton fused moe config for GLM-4.5-Air-FP8 on B200 by @zixuanzhang226 in https://github.com/vllm-project/vllm/pull/23695
* [Feature][Responses API] Support MCP tool in background mode by @wuhang2014 in https://github.com/vllm-project/vllm/pull/23494
* fix pynccl reduce_scatter by @youzhedian in https://github.com/vllm-project/vllm/pull/23648
* [quantization] use channel scales for w4a8 + misc fixes by @czhu-cohere in https://github.com/vllm-project/vllm/pull/23570
* [gpt-oss] Enable unit test for response API harmony integration by @heheda12345 in https://github.com/vllm-project/vllm/pull/23533
* [Bugfix] Lazy import gpt_oss_triton_kernels_moe for mxfp4 by @mgoin in https://github.com/vllm-project/vllm/pull/23678
* [Docs] Fix math rendering in docs by @hmellor in https://github.com/vllm-project/vllm/pull/23676
* [Bugfix][gpt-oss] passing the cache config in gpt-oss by @frank-wei in https://github.com/vllm-project/vllm/pull/23613
* [Bugfix]: Qwen3 Coder Tool Parser by @ranpox in https://github.com/vllm-project/vllm/pull/23099
* [Core] Asynchronous h2d in merge_multimodal_embeddings via pinned memory. by @huachenheli in https://github.com/vllm-project/vllm/pull/23686
* [Model] Add Ernie4.5 VL Model Support by @CSWYF3634076 in https://github.com/vllm-project/vllm/pull/22514
* [Frontend] Add --log-error-stack to print stack trace for error response by @heheda12345 in https://github.com/vllm-project/vllm/pull/22960
* [Frontend] Optimize beam search performance by limiting concurrency by @heheda12345 in https://github.com/vllm-project/vllm/pull/23599
* [Quantization] Expand compressed-tensors MoE matching logic to support NFP4 + FP8 MoEs by @dsikka in https://github.com/vllm-project/vllm/pull/22674
* [XPU] Add xpu torch.compile support by @jikunshang in https://github.com/vllm-project/vllm/pull/22609
* [CI/Build] Remove redundant LoRA model tests by @jeejeelee in https://github.com/vllm-project/vllm/pull/23706
* [Bugfix] fix when config.yaml config value is list parse error by @lengrongfu in https://github.com/vllm-project/vllm/pull/23528
* [Core] Use key-only cache for `BaseMultiModalProcessor` by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23018
* [XPU]fix cuda event used in XPU model runner by @jikunshang in https://github.com/vllm-project/vllm/pull/23708
* [CI/Build] Remove redundant register in model init tests by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23715
* [Docs] Fix an admonition important by @windsonsea in https://github.com/vllm-project/vllm/pull/23726
* Optimize input preparation for FlashInfer [2/N] by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23174
* [Misc] Move CpuGpuBuffer to vllm/v1/utils.py by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23728
* [FlashInfer] Cache hyper params in metadata builder by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23732
* [CI/Build] Reduce LoRA layer test cases by @jeejeelee in https://github.com/vllm-project/vllm/pull/23721
* [XPU] Fix OOM issue for data parallel with Ray backend by @faaany in https://github.com/vllm-project/vllm/pull/22500
* [Docs] Fix a 1-2-3 list and style issues in tpu.md by @windsonsea in https://github.com/vllm-project/vllm/pull/23729
* [model] Support MiniCPM-V 4.5 by @tc-mb in https://github.com/vllm-project/vllm/pull/23586
* [Bugfix] Fix task field initialization when PYTHONOPTIMIZE is enabled by @cndoit18 in https://github.com/vllm-project/vllm/pull/23718
* [Misc] Remove unnecessary `_send_reconfig_message()` in `core_client.py` by @njhill in https://github.com/vllm-project/vllm/pull/23127
* [V1] [Hybrid] Disable prefix caching by default for hybrid or mamba-based models  by @tdoublep in https://github.com/vllm-project/vllm/pull/23716
* [Model] Explicit `default_pooling_type` interface by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23736
* Add vLLM Korea Meetup in the README.md and meetups.md by @rebel-hongseok in https://github.com/vllm-project/vllm/pull/23746
* Fix pre-commit on main by @hmellor in https://github.com/vllm-project/vllm/pull/23747
* [Model] Interface to enable batch-level DP support by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23733
* Only run `get_attr_docs` if generating help text by @hmellor in https://github.com/vllm-project/vllm/pull/23723
* [Feature] Add Hopper DeepGEMM E8M0 for DeepSeekV3.1 scale_fmt by @yewentao256 in https://github.com/vllm-project/vllm/pull/23666
* [Model] Enable native HF format InternVL support by @Isotr0py in https://github.com/vllm-project/vllm/pull/23742
* [Doc]: upgrade version of crate-ci tool for improved typo detection by @didier-durand in https://github.com/vllm-project/vllm/pull/23755
* [LogitsProcs] Deduplicate built-in LP implementation logic by @njhill in https://github.com/vllm-project/vllm/pull/23362
* [Docs] Remove in-tree Gaudi install instructions by @hmellor in https://github.com/vllm-project/vllm/pull/23628
* [BugFix] Fix topk_softmax assert by @ProExpertProg in https://github.com/vllm-project/vllm/pull/19764
* [Model] Merge `SupportsMultiModalWithRawInput` with `SupportsMultiModal` by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23749
* [V1] [Hybrid] Enable compile and piecewise CUDA graph for MiniMax-Text models by @tdoublep in https://github.com/vllm-project/vllm/pull/22589
* [Docs] Fix warnings in `mkdocs build` (continued) by @Zerohertz in https://github.com/vllm-project/vllm/pull/23743
* ci: Add arm64 docker build to release pipeline by @seemethere in https://github.com/vllm-project/vllm/pull/23210
* Disable `torch.compile` for dynamic rope models in Transformers backend by @hmellor in https://github.com/vllm-project/vllm/pull/23738
* [Multimodal] Generate mm_hash based on request metadata when caching is turned off by @ywang96 in https://github.com/vllm-project/vllm/pull/23690
* [V1][Mamba] - Enable V1 by default for Mamba Models by @Josephasafg in https://github.com/vllm-project/vllm/pull/23650
* DP/EP Support for gpt-oss with deepep-ht comm kernel on SM100 by @zyongye in https://github.com/vllm-project/vllm/pull/23608
* [Bugfix] Fix Marlin NVFP4 for modelopt by @mgoin in https://github.com/vllm-project/vllm/pull/23659
* [Feature] Add `VLLM_DISABLE_PAD_FOR_CUDAGRAPH` to Avoid Hang Issue by @yewentao256 in https://github.com/vllm-project/vllm/pull/23595
* [Bugfix] Fix for V1 priority scheduling crashes at preemption by @Hanchenli in https://github.com/vllm-project/vllm/pull/23713
* Migrate Qwen inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23473
* [Feature] models: pass layer prefix to replace_linear_class for per-layer quantization routing. Addresses #23239 by @Shrey1306 in https://github.com/vllm-project/vllm/pull/23556
* [Perf] Tune configs for triton block fp8 gemm H100/H200 by @mgoin in https://github.com/vllm-project/vllm/pull/23748
* Gracefully handle edge cases in harmony utils by @Ithanil in https://github.com/vllm-project/vllm/pull/23155
* [CI] make all multi-gpu weight loading tests run nightly by @killershrimp in https://github.com/vllm-project/vllm/pull/23792
* Add deprecation warning for lora_extra_vocab_size by @ahengljh in https://github.com/vllm-project/vllm/pull/23635
* [Transform] [Quantization] Add transforms to compressed tensors by @kylesayrs in https://github.com/vllm-project/vllm/pull/22486
* [CI] enable idefics3 and fuyu-8b test in multimodal test by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23790
* [Bugfix] when set offline model running error by @lengrongfu in https://github.com/vllm-project/vllm/pull/23711
* [Kernel] cuda kernels for upcoming decode context parallel feature by @youzhedian in https://github.com/vllm-project/vllm/pull/23791
* [New Model]: Support GteNewModelForSequenceClassification by @noooop in https://github.com/vllm-project/vllm/pull/23524
* [Model] Add PP support and VLM backbone compatability for GPT-OSS by @Isotr0py in https://github.com/vllm-project/vllm/pull/23680
* [FIXBUG] Add return_success parameter to moe_wna16_weight_loader function by @JartX in https://github.com/vllm-project/vllm/pull/22797
* [Doc]: fix typos in .md files (including those of #23751) by @didier-durand in https://github.com/vllm-project/vllm/pull/23825
* [CI/Build][Bugfix] Fix Qwen VL tests on CPU by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/23818
* [BugFix][Spec Decode] Use float64 for uniform_probs by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23803
* [Model] [gpt-oss] fix gpt-oss pp support by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23815
* [Doc]: fix typos in Python scripts by @didier-durand in https://github.com/vllm-project/vllm/pull/23828
* [Bugfix] Fix benchmark_moe.py for blockwise fp8. by @crischeng in https://github.com/vllm-project/vllm/pull/23823
* [CI] Fix linting error on main by @tdoublep in https://github.com/vllm-project/vllm/pull/23835
* [Model][gpt-oss] Support DP+EP for GPT-OSS with FlashInfer trtllm-gen MoE by @nvpohanh in https://github.com/vllm-project/vllm/pull/23819
* [Bugfix] Add fake mode around passes by @angelayi in https://github.com/vllm-project/vllm/pull/23349
* [ci] breaks down V1 Test into 3 groups of approx 30 minutes runtime by @jeanschmidt in https://github.com/vllm-project/vllm/pull/23757
* Add scale_config.yml file for Meta autoscalers for GH Actions by @jeanschmidt in https://github.com/vllm-project/vllm/pull/23840
* Migrate Llama4ImagePatchInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/22021
* [ROCm][Aiter] Add triton fp8 bmm kernel for mla by @divakar-amd in https://github.com/vllm-project/vllm/pull/23264
* [bugfix] [spec-decoding] fix data race in sample_recovered_tokens_kernel (vLLM v1) by @He-Jingkai in https://github.com/vllm-project/vllm/pull/23829
* [NVIDIA] Support SiluMul + NVFP4 quant fusion by @elvischenv in https://github.com/vllm-project/vllm/pull/23671
* chore: build release image by default by @simon-mo in https://github.com/vllm-project/vllm/pull/23852
* [BugFix][FlashInfer] Fix potential race condition for paged_kv_indptr_cpu by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23737
* [V1] Enable prefill optimization for Gemma3n by @sarckk in https://github.com/vllm-project/vllm/pull/22628
* [Log] Use Debug Once for DeepGEMM E8M0 When not Enabled by @yewentao256 in https://github.com/vllm-project/vllm/pull/23858
* [V0 Deprecation] Remove V0 Samplers test by @WoosukKwon in https://github.com/vllm-project/vllm/pull/23862
* [XPU] support data parallel for MoE models on XPU by @chaojun-zhang in https://github.com/vllm-project/vllm/pull/22887
* [Models] Improve iteration over layers by @lgeiger in https://github.com/vllm-project/vllm/pull/19497
* [ROCm][Fix] Fix rocm build caused by #23791 by @charlifu in https://github.com/vllm-project/vllm/pull/23847
* [tests] Improve speed and reliability of test_transcription_api_correctness by @russellb in https://github.com/vllm-project/vllm/pull/23854
* [Bugfix] Use `ReplicatedLinear` for SequenceClassification head by @Isotr0py in https://github.com/vllm-project/vllm/pull/23836
* [BugFix][AMD][Deepseek] fix a dtype mismatch error for deepseek running on AMD by @KingsleyZhang123 in https://github.com/vllm-project/vllm/pull/23864
* [Platform] import activation_quant_fusion for CUDA only by @wangxiyuan in https://github.com/vllm-project/vllm/pull/23882
* Fix(async): Add support for truncate_prompt_tokens in AsyncLLM by @oneraghavan in https://github.com/vllm-project/vllm/pull/23800
* [CI/Build] Clean up LoRA test by @jeejeelee in https://github.com/vllm-project/vllm/pull/23890
* [mrope][Qwen2-VL] Fix edge case where getting index of image/video token can potentially throw in default vl mrope implementation.  by @huachenheli in https://github.com/vllm-project/vllm/pull/23895
* [Misc] Fix warnings for mistral model by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23552
* Better errors for Transformers backend missing features by @hmellor in https://github.com/vllm-project/vllm/pull/23759
* [V0 Deprecation] Remove pooling model support in V0  by @maxdebayser in https://github.com/vllm-project/vllm/pull/23434
* [CPU] Enable data parallel for CPU backend by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/23903
* [Performance] V1 Classify Models E2E Performance Optimization by @noooop in https://github.com/vllm-project/vllm/pull/23541
* [Multimodal] Consolidate mm inputs into MultiModalFeatureSpec by @sfeng33 in https://github.com/vllm-project/vllm/pull/23779
* Update PyTorch to 2.8.0 by @huydhn in https://github.com/vllm-project/vllm/pull/20358
* Adds `json_count_leaves` utility function  by @aditchawdhary in https://github.com/vllm-project/vllm/pull/23899
* [MODEL] `Apertus` and `XIELU` by @EduardDurech in https://github.com/vllm-project/vllm/pull/23068
* [Models] Use in-place adds in Idefics2Vision by @lgeiger in https://github.com/vllm-project/vllm/pull/23932
* [BugFix] Async scheduling and PP compatibility with DP by @njhill in https://github.com/vllm-project/vllm/pull/23770
* [CI]  Add `aiter` to matching list of issue auto labeller for `rocm` tag by @vllmellm in https://github.com/vllm-project/vllm/pull/23942
* [BUGFIX ] fix undefined silu_and_mul_nvfp4_quant by @youzhedian in https://github.com/vllm-project/vllm/pull/23929
* [RL][BugFix] Fix missing tokenizer error for token-in-token-out by @22quinn in https://github.com/vllm-project/vllm/pull/23904
* Tuned H100/H200 triton fp8 block configs for fused_qkv_a_proj by @mgoin in https://github.com/vllm-project/vllm/pull/23939
* [Docs] [V1] [Hybrid] Add new documentation re: contributing mamba-based models  by @tdoublep in https://github.com/vllm-project/vllm/pull/23824
* Revert gemma3n fast prefill changes by @sarckk in https://github.com/vllm-project/vllm/pull/23897
* [Misc] Make `download_weights_from_hf` more reliable by @hmellor in https://github.com/vllm-project/vllm/pull/23863
* [CI] Fix unavailable image remote URL by @ywang96 in https://github.com/vllm-project/vllm/pull/23966
* [Bugfix] Fix --config arg expansion called from api_server.py by @dubejf in https://github.com/vllm-project/vllm/pull/23944
* Add routed_scaling_factor to MoE grouped topk by @xyang16 in https://github.com/vllm-project/vllm/pull/23123
* [CI] Move testing image from remote URL to S3 by @ywang96 in https://github.com/vllm-project/vllm/pull/23980
* [CI] Fix broken compile tests due to unsupported SiluMul+Nvfp4Quant fusion by @sarckk in https://github.com/vllm-project/vllm/pull/23973
* [Core] Cleanup TPU model runner for MM by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/23894
* [V1] [Hybrid] Move MiniMaxLinearAttention into layers/mamba by @tdoublep in https://github.com/vllm-project/vllm/pull/23831
* [Bugfix] Fix test_lora_resolvers.py by @jeejeelee in https://github.com/vllm-project/vllm/pull/23984
* [UT] fix unify_kv_cache_configs when kv cache config needs sort by @andyxning in https://github.com/vllm-project/vllm/pull/23843
* [Model] Enable encoder DP for MiniCPM-V by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23948
* Add LoRA support for DeepSeek models (V2, V3, R1-0528) by @sadeghja1070 in https://github.com/vllm-project/vllm/pull/23971
* [Misc] add reorder_batch AttentionMetadataBuilder by @andyxning in https://github.com/vllm-project/vllm/pull/23798
* [Refactor] refactor freezing_value/cuda_event initialize outside try finally by @andyxning in https://github.com/vllm-project/vllm/pull/23758
* [Misc] enhance type hint for rearrange return value by @andyxning in https://github.com/vllm-project/vllm/pull/23519
* [LoRA] Much faster startup when LoRA is enabled by @andylolu2 in https://github.com/vllm-project/vllm/pull/23777
* Fix wrong truncate_prompt_tokens type hint by @gmarinho2 in https://github.com/vllm-project/vllm/pull/22761
* [Core][Multimodal] Allow passing `multi_modal_uuids` as multimodal identifiers. by @ywang96 in https://github.com/vllm-project/vllm/pull/23394
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24001
* vllm fix check on max vocab size by @xw285cornell in https://github.com/vllm-project/vllm/pull/22471
* [Minor] Fix some random typos in comments by @njhill in https://github.com/vllm-project/vllm/pull/24009
* v1: Support KV events from connectors by @orozery in https://github.com/vllm-project/vllm/pull/19737
* [BUGFIX] GPTQ quantization compatibility for Qwen3 MOE models (AutoGPTQ and AutoRound-GPTQ) by @JartX in https://github.com/vllm-project/vllm/pull/23994
* [Misc] Avoid redundant copy for encoder-only models by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24012
* Fix the bug related to loading GPTP INT3 weights. by @Jun-Howie in https://github.com/vllm-project/vllm/pull/23328
* [Misc] Move fast prefill logic to separate method by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24013
* [CI/Build] Improve Tensor Schema tests speed by avoid engine core initialization by @Isotr0py in https://github.com/vllm-project/vllm/pull/23357
* [Misc] refactor code by import as for torch._inductor.config by @andyxning in https://github.com/vllm-project/vllm/pull/23677
* Migrate Phi4 inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23471
* [Misc] IO Processor plugins for pooling models by @christian-pinto in https://github.com/vllm-project/vllm/pull/22820
* [Bugfix] Add support for `<tool_call>` format in streaming mode for XLAM Tool Parser by @DevonPeroutky in https://github.com/vllm-project/vllm/pull/22769
* [Misc] add hash_function doc string by @andyxning in https://github.com/vllm-project/vllm/pull/24014
* [Misc] Enable V1 FP16 inference on pre-Ampere GPUs by @Isotr0py in https://github.com/vllm-project/vllm/pull/24022
* [Frontend] Update the warning log when using VLLM_ALLOW_LONG_MAX_MODEL_LEN by @noooop in https://github.com/vllm-project/vllm/pull/20904
* [Kernel] Update DeepGEMM to latest commit by @jeejeelee in https://github.com/vllm-project/vllm/pull/23915
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24026
* [Frontend] Gemma3n audio `transcriptions`/`translations` endpoint by @NickLucche in https://github.com/vllm-project/vllm/pull/23735
* [Doc]: Fix CPU install docs: force torch-backend=cpu to avoid GPU torchvision errors by @yankay in https://github.com/vllm-project/vllm/pull/24033
* [Model]: support KeyeVL-1_5-8B by @Kwai-Keye in https://github.com/vllm-project/vllm/pull/23838
* Document multi-proc method selection for profiling by @hypdeb in https://github.com/vllm-project/vllm/pull/23802
* [Misc] Minor code simplification for spec decode by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24053
* [docs][misc] IOProcessor plugins fixes by @christian-pinto in https://github.com/vllm-project/vllm/pull/24046
* [Model] Support DP for ViT on Kimi-VL-A3B-Thinking-2506 by @david6666666 in https://github.com/vllm-project/vllm/pull/23817
* [Chore][V0 Deprecation] Move LogProb to a separate file by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24055
* [bugfix]fix MTP hidden states by @luccafong in https://github.com/vllm-project/vllm/pull/24056
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24042
* [V1][Mamba1] - FP32 SSM Kernel Support by @Josephasafg in https://github.com/vllm-project/vllm/pull/23506
* [Bugfix] Fix the issue that Blip2ForConditionalGeneration' object has… by @DamonJiang777 in https://github.com/vllm-project/vllm/pull/24028
* Remove runtime checks based on pooling params by @maxdebayser in https://github.com/vllm-project/vllm/pull/24051
* Migrate OvisImagePatchInputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/22024
* [XPU][Feature] fp8 online quantization support for XPU by @yma11 in https://github.com/vllm-project/vllm/pull/23148
* Migrate Interns1 inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23510
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24077
* [Model] Support dp on ViT on GLM-4.5V by @david6666666 in https://github.com/vllm-project/vllm/pull/23168
* [CI]: reduce HTTP calls inside entrypoints openai tests by @AzizCode92 in https://github.com/vllm-project/vllm/pull/23646
* correct LWS deployment yaml by @cberge908 in https://github.com/vllm-project/vllm/pull/23104
* [Gemma3n] Fix audio batching by @NickLucche in https://github.com/vllm-project/vllm/pull/24052
* [BugFix] Fix EXAONE4 rotary embeddings by @lkm2835 in https://github.com/vllm-project/vllm/pull/23918
* [Model] Classification models support logit_bias / sigmoid_normalize by @noooop in https://github.com/vllm-project/vllm/pull/24031
* [CI Failure] Skip failing nvfp4 silu test by @mgoin in https://github.com/vllm-project/vllm/pull/23959
* [docs] add SYS_NICE cap & `security-opt` for docker/k8s by @panpan0000 in https://github.com/vllm-project/vllm/pull/24017
* [Benchmark] Add support for local hf dataset path in benchmark by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23999
* [Bugfix] Fix transform_config parsing in Compressed Tensors by @kylesayrs in https://github.com/vllm-project/vllm/pull/23945
* Run ruff format on a few files. by @huachenheli in https://github.com/vllm-project/vllm/pull/24075
* [Bugfix] Fix packed_factor missing attribute error by @kyuyeunk in https://github.com/vllm-project/vllm/pull/23902
* [Metrics] Deprecate TPOT in favor of ITL by @markmc in https://github.com/vllm-project/vllm/pull/24110
* Fix weights loading for Apertus by @nathanrchn in https://github.com/vllm-project/vllm/pull/24100
* [Log] Only Print Profiler Results on Rank 0 by @yewentao256 in https://github.com/vllm-project/vllm/pull/23370
* [CI] Enable all hf transformers baselines in test_hybrid by @tdoublep in https://github.com/vllm-project/vllm/pull/23936
* [AMD][Kernel][Bugfix] Cast offsets tensor bn to tl.int64 to avoid GPU segfault by @rasmith in https://github.com/vllm-project/vllm/pull/23692
* [Bug] R1 Accuracy: Fix `routed_scaling_factor` Double Mul Issue by @yewentao256 in https://github.com/vllm-project/vllm/pull/24119
* [CI/Build] Disable SiluMul NVFP4 quant fusion tests by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/24121
* [XPU] Fix the bug of LoRA logits on the XPU platform by @chaojun-zhang in https://github.com/vllm-project/vllm/pull/24081
* Update release pipeline post PyTorch 2.8.0 update by @youkaichao in https://github.com/vllm-project/vllm/pull/24073
* Upgrade xgrammar to 0.1.23 by @russellb in https://github.com/vllm-project/vllm/pull/22988
* [V1] Wrapper which plumbs request-level logits processors into vLLM batch-level logits processing by @afeldman-nm in https://github.com/vllm-project/vllm/pull/23656
* fix some typos by @co63oc in https://github.com/vllm-project/vllm/pull/24071
* [Compile] Fix Compile Warning for `w4a8_mm_entry.cu` by @yewentao256 in https://github.com/vllm-project/vllm/pull/23660
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24093
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24115
* [Misc] Add check for dual_chunk_attention by @ZJY0516 in https://github.com/vllm-project/vllm/pull/24070
* [BugFix] Fix routed_scaling_factor double mul for dots1 and glm4 MoE models by @sarckk in https://github.com/vllm-project/vllm/pull/24132
* [distributed][rl] remove nccl cumem env var override by @youkaichao in https://github.com/vllm-project/vllm/pull/24141
* [Nixl] Heterogeneous TP support FlashInfer by @NickLucche in https://github.com/vllm-project/vllm/pull/20189
* [CI/Build] Serve images used by multimodal tests through local HTTP Server by @divyanshsinghvi in https://github.com/vllm-project/vllm/pull/23907
* [Misc] Clean up deadcode for legacy processing pipeline by @Isotr0py in https://github.com/vllm-project/vllm/pull/24153
* [CI] Accelerate mteb test by setting SentenceTransformers mteb score to a constant by @noooop in https://github.com/vllm-project/vllm/pull/24088
* Support add_generation_prompt in embeddings endpoint with chat request by @biba10 in https://github.com/vllm-project/vllm/pull/23931
* Fix MiniMax attention module prefix and remove useless code by @qscqesze in https://github.com/vllm-project/vllm/pull/23982
* FIX: Add libnuma-dev to Dockerfile for dev stage by @dongbo910220 in https://github.com/vllm-project/vllm/pull/20388
* [Bugfix] Fixing division by zero in triton_attn if query_heads/kv_heads > 16  by @bringlein in https://github.com/vllm-project/vllm/pull/23424
* [V1] v1 engine + full CUDA graph support for PLaMo2 by @nopperl in https://github.com/vllm-project/vllm/pull/23998
* [Kernels] Overlap shared experts with send/recv by @bnellnm in https://github.com/vllm-project/vllm/pull/23273
* Migrate whisper inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23505
* [Attention] Blackwell FP8 MLA support with CUTLASS_MLA backend by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/23289
* [Feature][P/D]: Optimize NIXL Connector xfer Launch by @david6666666 in https://github.com/vllm-project/vllm/pull/23887
* [Bugfix][DP] DP distribution does not require ray[default] by @kebe7jun in https://github.com/vllm-project/vllm/pull/23822
* [Feature][gpt-oss] Add support for num_cached_tokens and num_reasoning_tokens tracking by @NagyGeorge in https://github.com/vllm-project/vllm/pull/23460
* Remove deprecated `PyNcclConnector` by @panpan0000 in https://github.com/vllm-project/vllm/pull/24151
* [Feature][Responses API]Support MCP tools with streaming mode + background mode by @wuhang2014 in https://github.com/vllm-project/vllm/pull/23927
* [Kernel][Bugfix] Fix grouped topk cu by @mayuyuace in https://github.com/vllm-project/vllm/pull/24146
* [Refactor] Introduce basic Renderer for completion-style request by @sfeng33 in https://github.com/vllm-project/vllm/pull/24010
* Migrate ultravox inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23503
* [CPU] Refactor CPU unquantized linear by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/24150
* [Misc] Enhance output readability of helper script by @wdhongtw in https://github.com/vllm-project/vllm/pull/24214
* [Model] Add MiDashengLM model support by @bingchen-mi in https://github.com/vllm-project/vllm/pull/23652
* [Core][Model] Terratorch backend integration by @mgazz in https://github.com/vllm-project/vllm/pull/23513
* Improve flexibility of auto_tune.sh execution. by @anthonsu in https://github.com/vllm-project/vllm/pull/23766
* [Attention][Platform] Refactor MLA to support Custom Op by @whx-sjtu in https://github.com/vllm-project/vllm/pull/23332
* [Bugfix] Fix Incremental Detokenization with `tokenizers == 0.22.0` by @faaany in https://github.com/vllm-project/vllm/pull/24159
* [Attention] FlashAttn MLA by @LucasWilkinson in https://github.com/vllm-project/vllm/pull/14258
* [Hardware][Apple-CPU] Disable OneDNN build for Apple Silicon by @ignaciosica in https://github.com/vllm-project/vllm/pull/24200
* [Feature][Response API] Add streaming support for non-harmony by @kebe7jun in https://github.com/vllm-project/vllm/pull/23741
* [Doc] Update vLLM Singapore Meetup info by @tjtanaa in https://github.com/vllm-project/vllm/pull/24234
* [Model] Add pp support for hunyuan by @ZJY0516 in https://github.com/vllm-project/vllm/pull/24212
* Use hidden_size_per_head as head_size fallback by @nopperl in https://github.com/vllm-project/vllm/pull/24221
* [XPU] support Triton Attention backend on Intel GPU by @jikunshang in https://github.com/vllm-project/vllm/pull/24149
* [LoRA]: Add lora support to qwen-2.5-omni by @pratapyash in https://github.com/vllm-project/vllm/pull/24231
* [Misc] Removed force_fp8_e4m3fnuz from FP8LinearOp by @nvjullin in https://github.com/vllm-project/vllm/pull/23725
* [Perf] Freeze core engine proc heap after init by @njhill in https://github.com/vllm-project/vllm/pull/24008
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24173
* [Misc] Slight improve deepgemm print by @jeejeelee in https://github.com/vllm-project/vllm/pull/24085
* Upgrade FlashInfer to v0.3.0 by @nvpohanh in https://github.com/vllm-project/vllm/pull/24086
* QWEN3 Coder Fused MoE kernels Optimization configs by @samanamp in https://github.com/vllm-project/vllm/pull/24266
* [Misc] Have AsyncLLM `custom_stat_loggers` extend default logger list by @eicherseiji in https://github.com/vllm-project/vllm/pull/20952
* [Bugfix][Misc] Fix silu_and_mul_nvfp4_quant issue and extract common utils for nvfp4 kernel source files by @elvischenv in https://github.com/vllm-project/vllm/pull/23727
* [CI/Build] Reduce the number of redundant cases to test for LoRA by @zhuohan123 in https://github.com/vllm-project/vllm/pull/24276
* [Frontend] Skip unnecessary detokenization when token_id is requested by @NickLucche in https://github.com/vllm-project/vllm/pull/24236
* [gpt-oss] tool parser supports for /chat/completions [1/n] by @aarnphm in https://github.com/vllm-project/vllm/pull/22386
* [XPU][P/D] Add XPU support in NixlConnector by @zhenwei-intel in https://github.com/vllm-project/vllm/pull/22436
* Adding int4 and int8 models for CPU benchmarking by @louie-tsai in https://github.com/vllm-project/vllm/pull/23709
* [docs] add shenzhen meetup by @youkaichao in https://github.com/vllm-project/vllm/pull/24326
* [gpt-oss][Bugfix]Fix streamableparser for missing handling of certain token_ids by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/24306
* [Bugfix] Fix silu_mul+quant fusion test by @elvischenv in https://github.com/vllm-project/vllm/pull/24341
* [RFC] allow cancelation after shutdown in blocking collective_rpc by @842974287 in https://github.com/vllm-project/vllm/pull/23390
* [CI] Add timeouts to tests by @rafvasq in https://github.com/vllm-project/vllm/pull/24260
* [Perf][V1] Fully overlap model execution by @benchislett in https://github.com/vllm-project/vllm/pull/23569
* Add @22quinn as code reviewer for RL related components by @22quinn in https://github.com/vllm-project/vllm/pull/24346
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24294
* [KV Sharing] Raise error if using eagle with fast prefill by @sarckk in https://github.com/vllm-project/vllm/pull/24350
* [Feature] Support Decode Context Parallel (DCP) for MLA by @youzhedian in https://github.com/vllm-project/vllm/pull/23734
* [Bugfix] Catch and log invalid token ids in detokenizer by @njhill in https://github.com/vllm-project/vllm/pull/24351
* [Core] Allow disabling TP sharding for parallel Linear layer by @Isotr0py in https://github.com/vllm-project/vllm/pull/23024
* [New Model]: google/embeddinggemma-300m by @noooop in https://github.com/vllm-project/vllm/pull/24318
* refactor: Turn GPUModelRunner.inputs_embeds to a CpuGpuBuffer by @qthequartermasterman in https://github.com/vllm-project/vllm/pull/24345
* [Multimodal] Improve max video embedding length estimation in V1 by @ywang96 in https://github.com/vllm-project/vllm/pull/24312
* [CI] Disable flaky structured output test from CI by @ywang96 in https://github.com/vllm-project/vllm/pull/24366
* Add @benchislett to codeowner for spec decode and structured outputs by @benchislett in https://github.com/vllm-project/vllm/pull/24362
* [Bugfix] Avoid uninitialized usage of azp_val when AZP is false. by @mohankku in https://github.com/vllm-project/vllm/pull/24335
* [Bugfix] Fix broken deepseek fp8 TP weights loading by @Isotr0py in https://github.com/vllm-project/vllm/pull/24367
* [Bugfix] Fix test_mixtral_moe by @jeejeelee in https://github.com/vllm-project/vllm/pull/24371
* Lora bias(enable_lora_bias) deprecate warning by @ashwin-phadke in https://github.com/vllm-project/vllm/pull/24339
* [Fix] [gpt-oss] fix non-tool calling path for chat completion by @aarnphm in https://github.com/vllm-project/vllm/pull/24324
* [Frontend][Responses API] Support reporting tool output tokens and fix reasoning token count by @yeqcharlotte in https://github.com/vllm-project/vllm/pull/24285
* [Bugfix] Fix unstable silu_mul+nvfp4 quant fusion test by @elvischenv in https://github.com/vllm-project/vllm/pull/24370
* break execute_model in gpu_model_runner into sub-functions for custom scopes by @bangshengtang in https://github.com/vllm-project/vllm/pull/24265
* [V0 deprecation] Deprecate V0 Neuron backend by @WoosukKwon in https://github.com/vllm-project/vllm/pull/21159
* [attention][DCP] use AttentionImpl.need_to_return_lse_for_decode by @youkaichao in https://github.com/vllm-project/vllm/pull/24372
* Migrate Qwen2 inputs to TensorSchema by @bbeckca in https://github.com/vllm-project/vllm/pull/23475
* [CI][Fix] deterministic seed for flaky CI runs on structured outputs by @aarnphm in https://github.com/vllm-project/vllm/pull/24380
* [Benchmark] add benchmark for custom activation op by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23908
* QWEN3 Thinking Fused MoE kernels Optimization configs by @samanamp in https://github.com/vllm-project/vllm/pull/24330
* [Misc] collect flashinfer version in collect_env.py by @yeqcharlotte in https://github.com/vllm-project/vllm/pull/24378
* [Bugfix] Fix Qwen3-coder moe tuned config by @jeejeelee in https://github.com/vllm-project/vllm/pull/24072
* [TPU] Remove TopKTopPSampler dependency for TPU sampler by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24391
* Add renderer-based prompt processing for embedding and classification endpoints by @sfeng33 in https://github.com/vllm-project/vllm/pull/24356
* Skip MM Encoder for non-first PP ranks by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24387
* Add @luccafong to codeowner for spec decode by @luccafong in https://github.com/vllm-project/vllm/pull/24397
* [Kernel] Support decode context parallelism on Blackwell with CUTLASS MLA by @minosfuture in https://github.com/vllm-project/vllm/pull/24385
* [xpu] upgrade ipex/python3.12 for xpu by @yma11 in https://github.com/vllm-project/vllm/pull/23830
* [Sampler] Support returning all prompt logprobs by @charlotte12l in https://github.com/vllm-project/vllm/pull/23868
* [CI/Build] Disable flaky test_structured_output tests by @22quinn in https://github.com/vllm-project/vllm/pull/24404
* [CI/Build] Fix local image inputs in test_pixtral.py by @huachenheli in https://github.com/vllm-project/vllm/pull/24401
* [Doc] Fix UTF-8 encoding issues in documentation generation on Windows by @alhridoy in https://github.com/vllm-project/vllm/pull/24361
* [P/D] Add a shutdown method to the Connector API by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/22699
* [Model] Remove unnecessary CUDA sync of GLM-4.1V image and video preprocess by @what-in-the-nim in https://github.com/vllm-project/vllm/pull/24332
* [Model] Remove unnecessary CUDA sync of Qwen2VL image and video preprocess by @what-in-the-nim in https://github.com/vllm-project/vllm/pull/24334
* [gpt-oss][Responses API] Fix the function call id format by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/24409
* [Docs] Fix a tip indentation and typo by @windsonsea in https://github.com/vllm-project/vllm/pull/24419
* [Doc]: fix typos in Python comments by @didier-durand in https://github.com/vllm-project/vllm/pull/24417
* [Doc] Fix issues in integrations/llamastack.md by @windsonsea in https://github.com/vllm-project/vllm/pull/24428
* [Bugfix] Fix get_quant_config when using modelscope by @Potabk in https://github.com/vllm-project/vllm/pull/24421
* [Bugfix] Fix mamba2 prefill chunking by @tomeras91 in https://github.com/vllm-project/vllm/pull/23279
* [Misc] Terratorch related fixes by @christian-pinto in https://github.com/vllm-project/vllm/pull/24337
* Move `KVEventsConfig` from `config/__init__.py` to `config/kv_events.py` by @hmellor in https://github.com/vllm-project/vllm/pull/24433
* [Frontend] User-provided uuids for medias in chat. (RFC #22044) by @huachenheli in https://github.com/vllm-project/vllm/pull/23449
* [Docs] Move feature compatibility tables to README by @hmellor in https://github.com/vllm-project/vllm/pull/24431
* [Doc]: fix 2 hyperlinks leading to Ray site after they changed Ray's doc structure by @didier-durand in https://github.com/vllm-project/vllm/pull/24438
* [Docs]add eplb_config param use docs by @lengrongfu in https://github.com/vllm-project/vllm/pull/24213
* [Model] Enable BNB support for qwen2_5_omni_thinker by @jeejeelee in https://github.com/vllm-project/vllm/pull/24420
* [Spec Decode][Benchmark] Add Spec Bench Dataset for benchmarking by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/23563
* [Spec Decode][Benchmark] Add Blitzedit dataset by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/23605
* [Model] Remove quantized mixtral by @jeejeelee in https://github.com/vllm-project/vllm/pull/24437
* [CI] Enable encoder model compilation test by @ZJY0516 in https://github.com/vllm-project/vllm/pull/24442
* [Model loader]: support multi-thread model weight loading by @BraveY in https://github.com/vllm-project/vllm/pull/23928
* [Spec Decode] Fix offline spec_decode.py by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/24257
* [Attention] FlashAttention MLA cudagraph support by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/23958
* [Bugfix] Disable the statslogger if the api_server_count is greater than 1 by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/22227
* [Hardware][IBM Z] Fix Outlines Core issue for s390x by @R3hankhan123 in https://github.com/vllm-project/vllm/pull/24034
* [CI] Add nightly multiarch manifests to dockerhub by @csahithi in https://github.com/vllm-project/vllm/pull/24102
* Update reviewers for modelopt related files by @Edwardf0t1 in https://github.com/vllm-project/vllm/pull/24468
* [Bugfix][Wide EP] Fix redundant work when using DeepEP, TP Attn, and EP MoE by @tlrmchlsmth in https://github.com/vllm-project/vllm/pull/24134
* [gpt-oss] Harmony changes with container tool support by @morgendave in https://github.com/vllm-project/vllm/pull/23386
* Bump actions/setup-python from 5.4.0 to 6.0.0 by @dependabot[bot] in https://github.com/vllm-project/vllm/pull/24414
* [doc] update `vllm serve` cli args documentation by @cjackal in https://github.com/vllm-project/vllm/pull/24329
* Bump actions/stale from 9.1.0 to 10.0.0 by @dependabot[bot] in https://github.com/vllm-project/vllm/pull/24412
* Bump actions/github-script from 7.0.1 to 8.0.0 by @dependabot[bot] in https://github.com/vllm-project/vllm/pull/24413
* Move `KVTransferConfig` from `config/__init__.py` to `config/kv_transfer.py` by @hmellor in https://github.com/vllm-project/vllm/pull/24434
* [BugFix][Model] Fix Ernie4.5-VL hanging on long inputs by @CSWYF3634076 in https://github.com/vllm-project/vllm/pull/24074
* [Flashinfer] Support Flashinfer TRTLLM FP8-qkv BF16/FP16-out Attention Kernel by @elvischenv in https://github.com/vllm-project/vllm/pull/23647
* [Core] Use sha256 bytes instead of BlockHash to reduce GC overhead by @linzebing in https://github.com/vllm-project/vllm/pull/23673
* Add data_parallel_size to VllmConfig string representation by @Prowindy in https://github.com/vllm-project/vllm/pull/24298
* [Bugfix] Fix Apertus HF repo name by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/24447
* [Misc] Improve Worker process title and logging prefix by @22quinn in https://github.com/vllm-project/vllm/pull/22205
* [Doc] mention fpdb for multiprocess breakpoints by @mickaelseznec in https://github.com/vllm-project/vllm/pull/24452
* [Misc] Support bench serve long context by @minosfuture in https://github.com/vllm-project/vllm/pull/24373
* [Doc]: fixing typos to improve docs by @didier-durand in https://github.com/vllm-project/vllm/pull/24480
* [Performance][MM] Building the inverse permutation in O(n) time in Qwen2_5_VisionTransformer by @david6666666 in https://github.com/vllm-project/vllm/pull/24443
* [Misc] Add claude settings to gitignore by @yeqcharlotte in https://github.com/vllm-project/vllm/pull/24492
* [Misc] Add Codex settings to gitignore by @ywang96 in https://github.com/vllm-project/vllm/pull/24493
* [gpt-oss] Validate gpt-oss python tool during initialization by @heheda12345 in https://github.com/vllm-project/vllm/pull/23856
* [RL] fast weight update with zmq + ipc handles by @weixiao-huang in https://github.com/vllm-project/vllm/pull/24295
* [CI/Build][Doc] Fully deprecate old bench scripts for serving / throughput / latency by @yeqcharlotte in https://github.com/vllm-project/vllm/pull/24411
* [Compilation][WideEP] Enable Piecewise CUDAGraph for DeepEPHT by @yewentao256 in https://github.com/vllm-project/vllm/pull/24123
* [Model] Systematic support for fp32 head, pooling models part by @noooop in https://github.com/vllm-project/vllm/pull/23810
* [Bugfix] Handle the edge case in detokenizer where processed tokens contain both `stop` str and `eos` token by @dtransposed in https://github.com/vllm-project/vllm/pull/23938
* [Core] Run garbage collector after CUDA graph capture to fix throughput regression by @micah-wil in https://github.com/vllm-project/vllm/pull/24128
* [Kernels] Add Flash Linear Attention Kernels by @youkaichao in https://github.com/vllm-project/vllm/pull/24518
* [ROCm][CI/Build] Sync ROCm dockerfiles with the ROCm fork by @gshtras in https://github.com/vllm-project/vllm/pull/24279
* [Bugfix] Fix  hidden_size for multimodal classification model by @jeejeelee in https://github.com/vllm-project/vllm/pull/24501
* Extend renderer with embedding support and integrate completion endpoint by @sfeng33 in https://github.com/vllm-project/vllm/pull/24405
* [Misc] bump outlines_core to fix the version conflicts with outlines >= 1.2.0 by @serihiro in https://github.com/vllm-project/vllm/pull/24368
* [Docs] Gemma3n `transcriptions` endpoint support by @NickLucche in https://github.com/vllm-project/vllm/pull/24512
* [TPU] Fix tpu structured decoding in mixed batches by @Chenyaaang in https://github.com/vllm-project/vllm/pull/24458
* [CI] execute all piecewise compilation tests together by @ZJY0516 in https://github.com/vllm-project/vllm/pull/24502
* [Feature] Disallow FlashMLA on Blackwell by @yewentao256 in https://github.com/vllm-project/vllm/pull/24521
* [Log] Use a relative path in debug-level logs to distinguish files with identical names by @ZJY0516 in https://github.com/vllm-project/vllm/pull/23846
* [Benchmark] Update bench doc with mtbench, blazedit, spec bench by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/24450
* [Benchmark] Add option to skip oversampling in benchmark by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/24457
* [ROCm][Feature] Enable Pipeline Parallelism with Ray Compiled Graph on ROCm by @charlifu in https://github.com/vllm-project/vllm/pull/24275
* [Bugfix] Improve EPLB config validation error message by @tlrmchlsmth in https://github.com/vllm-project/vllm/pull/24524
* [Bugfix] Fix for 24530. Fix naive all2all shared expert overlap. by @bnellnm in https://github.com/vllm-project/vllm/pull/24538
* [Perf] Convert np array to torch tensor to index into block table for attn chunking by @sarckk in https://github.com/vllm-project/vllm/pull/24474
* Add @heheda12345 to CODEOWNERS of KVCacheManager related code by @heheda12345 in https://github.com/vllm-project/vllm/pull/24546
* [CI] Retry flaky fp8 cutlass mla tests by @njhill in https://github.com/vllm-project/vllm/pull/24536
* [Hardware][Apple-CPU] Enable native bfloat16 on Apple Silicon (M2 and later) by @ignaciosica in https://github.com/vllm-project/vllm/pull/24129
* [BugFix] Fix async core engine client finalizer by @njhill in https://github.com/vllm-project/vllm/pull/24540
* [CI] Adjust threshold for flaky ngram spec decoding test by @njhill in https://github.com/vllm-project/vllm/pull/24528
* [KV Connector] More async support for `get_num_new_matched_tokens` by @ApostaC in https://github.com/vllm-project/vllm/pull/23620
* [P/D] MultiConnector supports shutdown by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/24425
* [BugFix][Spec Decode] Fix out-of-range index triggered by eagle3; re-enable test for LlamaForCausalLMEagle3 by @wwl2755 in https://github.com/vllm-project/vllm/pull/24392
* [gpt-oss] Cache permute indices for faster MXFP4 MoE layer loading by @frank-wei in https://github.com/vllm-project/vllm/pull/24154
* [Core] Simplify and unify mm uuid handling & auto-generated mm hash overrides processing.  by @huachenheli in https://github.com/vllm-project/vllm/pull/24271
* [Bugfix] Update Run:AI Model Streamer Loading Integration by @pwschuurman in https://github.com/vllm-project/vllm/pull/23845
* [Docs] Enable relative links in examples to function when rendered in the docs by @hmellor in https://github.com/vllm-project/vllm/pull/24041
* [docs] promo pytorch conf and ray summit by @simon-mo in https://github.com/vllm-project/vllm/pull/24562
* [Bugfix] Guard `_may_reorder_batch` for encoder-only models on CPU (#24319) by @comsky in https://github.com/vllm-project/vllm/pull/24348
* Consolidate rendering parameters into RenderConfig dataclass by @sfeng33 in https://github.com/vllm-project/vllm/pull/24543
* [Model] Limit CPU threads for image transformations in InternVL to reduce cpu contention. by @li-jinpeng in https://github.com/vllm-project/vllm/pull/24519
* [Attention] add DCP support for FLASH_ATTN_MLA backend by @LucasWilkinson in https://github.com/vllm-project/vllm/pull/24453
* [ROCm][Bugfix] Fix Aiter RMSNorm  by @vllmellm in https://github.com/vllm-project/vllm/pull/23412
* [Docs] Improve organisation of API Reference nav by @hmellor in https://github.com/vllm-project/vllm/pull/24569
* [Docs] Document the extra memory footprint overhead when using EPLB by @tlrmchlsmth in https://github.com/vllm-project/vllm/pull/24537
* Support for NemotronH Nano VLM by @danielafrimi in https://github.com/vllm-project/vllm/pull/23644
* Feature/vit attention unification# 23880 by @baonudesifeizhai in https://github.com/vllm-project/vllm/pull/23978
* [LoRA]: Add LoRA support to Mistral's Voxtral models by @pratapyash in https://github.com/vllm-project/vllm/pull/24517
* Move `LoadConfig` from `config/__init__.py` to `config/load.py` by @hmellor in https://github.com/vllm-project/vllm/pull/24566
* [BugFix][Multi Modal] Fix TensorSchema shape mismatch in Molmo by @wwl2755 in https://github.com/vllm-project/vllm/pull/24559
* [BugFix][easy] Fix flaky test test_gpt_oss_multi_turn_chat by @lacora in https://github.com/vllm-project/vllm/pull/24549
* [BugFix] Ensure integrity of reused CPU tensors during async scheduling by @njhill in https://github.com/vllm-project/vllm/pull/24527
* [CI/Build] split true unit tests to Entrypoints Unit Tests by @yeqcharlotte in https://github.com/vllm-project/vllm/pull/24418
* [rocm] enable torchao quantization for rocm by @draftbk in https://github.com/vllm-project/vllm/pull/24400
* [CI] Add PPL test for generation models by @noooop in https://github.com/vllm-project/vllm/pull/24485
* [CI/Build] bump timm dependency by @dtrifiro in https://github.com/vllm-project/vllm/pull/24189
* fix some typos by @co63oc in https://github.com/vllm-project/vllm/pull/24167
* Fix Auto_Round Quatization Loading on SM75 and Lower GPUs by @RoadToNowhereX in https://github.com/vllm-project/vllm/pull/24217
* [Docs] Fix warnings in `mkdocs build` (continued) by @Zerohertz in https://github.com/vllm-project/vllm/pull/24092
* [BugFix] `python collect_env.py` and `vllm collect-env` compatibility with uv venv by @yankay in https://github.com/vllm-project/vllm/pull/24066
* [Platform] Custom ops support for LMhead and LogitsProcessor by @zzhx1 in https://github.com/vllm-project/vllm/pull/23564
* [CI] Fix tensorizer test assertion by @pwschuurman in https://github.com/vllm-project/vllm/pull/24545
* [Core] Split LoRA layers by @jeejeelee in https://github.com/vllm-project/vllm/pull/24574
* [Doc] Add documentation for GLM-4.5 series models: tool-calling and reasoning parser by @WangErXiao in https://github.com/vllm-project/vllm/pull/24589
* [Logging] allow config logging stream by @842974287 in https://github.com/vllm-project/vllm/pull/24336
* [Bugfix] fix modelopt exclude_modules name mapping by @tomeras91 in https://github.com/vllm-project/vllm/pull/24178
* [Bugfix] Fix DeepEP config for DP4TP4 by @minosfuture in https://github.com/vllm-project/vllm/pull/23619
* [Core] Support configuration parsing plugin by @charlotte12l in https://github.com/vllm-project/vllm/pull/24277
* [Misc] update log level debug to warning when process port is used by by @lengrongfu in https://github.com/vllm-project/vllm/pull/24226
* [Bugfix] Enable FP8 KV cache for FlashInfer and Triton backend on non-sm100 GPUs by @gau-nernst in https://github.com/vllm-project/vllm/pull/24577
* [CI] Fail subprocess tests with root-cause error by @njhill in https://github.com/vllm-project/vllm/pull/23795
* [v1] Add Whisper model support (encoder-decoder) by @russellb in https://github.com/vllm-project/vllm/pull/21088
* [torch.compile][ROCm][V1] Enable attention output FP8 fusion for V1 attention backends by @gshtras in https://github.com/vllm-project/vllm/pull/19767
* [gpt-oss] raise error for flashinfer backend without trtllm by @heheda12345 in https://github.com/vllm-project/vllm/pull/24482
* [Perf] Warmup FlashInfer attention during startup by @mgoin in https://github.com/vllm-project/vllm/pull/23439
* [Kernel] Flashinfer MLA (trtllm-gen) decode kernel integration by @hjjq in https://github.com/vllm-project/vllm/pull/21078
* [Misc] Make timeout passable in init_distributed_environment by @jberkhahn in https://github.com/vllm-project/vllm/pull/24522
* [Models][Quantization] Add quantization configuration update in Voxtral model by @anmarques in https://github.com/vllm-project/vllm/pull/24122
* [distributed] update known issues by @youkaichao in https://github.com/vllm-project/vllm/pull/24624
* Add @chaunceyjiang to codeowner for reasoning Reasoning and Tool parser by @chaunceyjiang in https://github.com/vllm-project/vllm/pull/24406
* [Bug] [Spec Decode] Fix model_initialization test and mismatch in aux_hidden_layers by @wwl2755 in https://github.com/vllm-project/vllm/pull/24613
* [Ultravox] Fix Gemma instantiation, support quantization via --hf-overrides by @petersalas in https://github.com/vllm-project/vllm/pull/24131
* [Bugfix] Add missing VIT backend dispatch on CPU by @bigPYJ1151 in https://github.com/vllm-project/vllm/pull/24623
* [BugFix] Fix pipeline parallel by @njhill in https://github.com/vllm-project/vllm/pull/24621
* [Engine][Chore] use local variable and remove output var assignment by @GuyStone in https://github.com/vllm-project/vllm/pull/24554
* Kimi K2 Fused MoE kernels Optimization configs by @samanamp in https://github.com/vllm-project/vllm/pull/24597
* Enable --profile in 'vllm bench throughput' by @tomasruizt in https://github.com/vllm-project/vllm/pull/24575
* [Core] feat: Add --safetensors-load-strategy flag for faster safetensors loading from Lustre by @shengshiqi-google in https://github.com/vllm-project/vllm/pull/24469
* [Doc]: fixing doc typos by @didier-durand in https://github.com/vllm-project/vllm/pull/24635
* [Model] New model support for Motif-1-Tiny by @ca1207 in https://github.com/vllm-project/vllm/pull/23414
* Remove redundant all gather + split by @chenxi-yang in https://github.com/vllm-project/vllm/pull/23441
* [torchao] Support quantization configs using module swap by @jerryzh168 in https://github.com/vllm-project/vllm/pull/21982
* Add the support for the qwen3 next model (a hybrid attention model). by @sighingnow in https://github.com/vllm-project/vllm/pull/24526
* [Bugfix] Fix incorrect import of CacheConfig by @DarkLight1337 in https://github.com/vllm-project/vllm/pull/24631
* [Docs] Revise frameworks/anything-llm.md by @windsonsea in https://github.com/vllm-project/vllm/pull/24489
* [Docs] Update V1 doc to reflect whisper support by @russellb in https://github.com/vllm-project/vllm/pull/24606
* [Docs] Use 1-2-3 list for deploy steps in deployment/frameworks/ by @windsonsea in https://github.com/vllm-project/vllm/pull/24633
* [CI]Add transformers_utils to Async Engine, Inputs, Utils, Worker Test by @charlotte12l in https://github.com/vllm-project/vllm/pull/24615
* [Bugfix] Fix _synced_weight_loader by @kyuyeunk in https://github.com/vllm-project/vllm/pull/24565
* [CI] Split pooling from entrypoints Test by @noooop in https://github.com/vllm-project/vllm/pull/24632
* [Misc] Add @NickLucche to codeowners by @NickLucche in https://github.com/vllm-project/vllm/pull/24647
* [CI Failure] fix models/language/pooling/test_auto_prefix_cache_support.py by @noooop in https://github.com/vllm-project/vllm/pull/24636
* Fix typing for `safetensors_load_strategy` by @hmellor in https://github.com/vllm-project/vllm/pull/24641
* Move `LoRAConfig` from `config/__init__.py` to `config/lora.py` by @hmellor in https://github.com/vllm-project/vllm/pull/24644
* [XPU] add missing dependency tblib for XPU CI by @faaany in https://github.com/vllm-project/vllm/pull/24639
* [Docs] Fixes a typo in the qwen3next model name. by @sighingnow in https://github.com/vllm-project/vllm/pull/24654
* [build] add torch to tool.uv no-build-isolation-package by @youkaichao in https://github.com/vllm-project/vllm/pull/24303
* [Bench] Add qwen-next in benchmark_moe.py by @jeejeelee in https://github.com/vllm-project/vllm/pull/24661
* [CI] Split mteb test from Language Models Test by @noooop in https://github.com/vllm-project/vllm/pull/24634
* Allow users to specify kv cache memory size by @BoyuanFeng in https://github.com/vllm-project/vllm/pull/21489
* [HybridKVCache][Platform] Add support_hybrid_kv_cache for platform by @MengqingCao in https://github.com/vllm-project/vllm/pull/24646
* [Bugifx] Fix qwen-next packed_modules_mapping by @jeejeelee in https://github.com/vllm-project/vllm/pull/24656
* [Docs] Add transcription support to model by @NickLucche in https://github.com/vllm-project/vllm/pull/24664
* [Doc] Fix Markdown Pre-commit Error by @yewentao256 in https://github.com/vllm-project/vllm/pull/24670
* [Docs] Fix typos in EP deployment doc by @hmellor in https://github.com/vllm-project/vllm/pull/24669
* [VLM] Optimize GLM4.5-V-style video processing to only decode necessary frames by @Isotr0py in https://github.com/vllm-project/vllm/pull/24161
* [Kernels] Enable Torch Symmetric Memory All-Reduce By Default by @ilmarkov in https://github.com/vllm-project/vllm/pull/24111
* [Bugfix] Fix platform-specific routing in CustomOp implementations by @kzawora-intel in https://github.com/vllm-project/vllm/pull/24444
* Fix model name included in responses by @hmellor in https://github.com/vllm-project/vllm/pull/24663
* fix some typos by @co63oc in https://github.com/vllm-project/vllm/pull/24616
* [Docs] Fix formatting of transcription doc by @hmellor in https://github.com/vllm-project/vllm/pull/24676
* [VLM] Migrate remain DP-supported ViT models to use `disable_tp` by @Isotr0py in https://github.com/vllm-project/vllm/pull/24363
* [Ultravox] Use wrapped_model_config to instantiate inner model by @petersalas in https://github.com/vllm-project/vllm/pull/24679
* [Doc] Remove Useless Comments by @yewentao256 in https://github.com/vllm-project/vllm/pull/24687
* [Qwen3-Next] Add MoE Config for H200 by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24688
* [BugFix] Fix tokenize asyncio task leak by @njhill in https://github.com/vllm-project/vllm/pull/24677
* update spec decode metrics to use throughput by @qandrew in https://github.com/vllm-project/vllm/pull/24127
* [Kernel][B200] `mxfp4` fused cutlass moe by @djmmoss in https://github.com/vllm-project/vllm/pull/23696
* [flashinfer] [kernel] support for fp8 kv cache for trtllm prefill attention by @mxz297 in https://github.com/vllm-project/vllm/pull/24197
* [Bugfix] Set `VLLM_ALLREDUCE_USE_SYMM_MEM` default to False by @yewentao256 in https://github.com/vllm-project/vllm/pull/24696
* [Qwen3-Next] MoE configs for H200 TP=1,2,4 by @WoosukKwon in https://github.com/vllm-project/vllm/pull/24695
* [CI/Build] Add bc-linter to vLLM CI by @zhewenl in https://github.com/vllm-project/vllm/pull/21234
* [Qwen3-Next] Add B200 MoE configs for Qwen3-next by @vadiklyutiy in https://github.com/vllm-project/vllm/pull/24698
* [Bugfix][Attention] Fix FlashInfer MLA block size logic by @MatthewBonanni in https://github.com/vllm-project/vllm/pull/24692
* [Perf] Use upstream CUTLASS for SM90 Block FP8 kernel by @mgoin in https://github.com/vllm-project/vllm/pull/23280
* [Qwen3-Next] MOE configs for H100 TP4 by @heheda12345 in https://github.com/vllm-project/vllm/pull/24699
* [Doc] Clarify cudagraph capture size logic and default behavior in scheduler by @Zazzle516 in https://github.com/vllm-project/vllm/pull/18698
* [Bug] Fix Layer `weight_block_size` Assertion Issue by @yewentao256 in https://github.com/vllm-project/vllm/pull/24674
* [Startup] Make DeepGEMM warmup scale with max-num-batched-tokens by @LucasWilkinson in https://github.com/vllm-project/vllm/pull/24693
* [V1] feat:add engine v1 tracing by @RichardoMrMu in https://github.com/vllm-project/vllm/pull/20372
* [Bugfix] fixes the causal_conv1d_update kernel update non-speculative decoding cases by @sighingnow in https://github.com/vllm-project/vllm/pull/24680

## New Contributors
* @DoubleVII made their first contribution in https://github.com/vllm-project/vllm/pull/23058
* @carlory made their first contribution in https://github.com/vllm-project/vllm/pull/23090
* @nikheal2 made their first contribution in https://github.com/vllm-project/vllm/pull/22725
* @Tialo made their first contribution in https://github.com/vllm-project/vllm/pull/23172
* @myselvess made their first contribution in https://github.com/vllm-project/vllm/pull/23084
* @yiz-liu made their first contribution in https://github.com/vllm-project/vllm/pull/23169
* @ultmaster made their first contribution in https://github.com/vllm-project/vllm/pull/22587
* @KilJaeeun made their first contribution in https://github.com/vllm-project/vllm/pull/22790
* @wzshiming made their first contribution in https://github.com/vllm-project/vllm/pull/23242
* @misrasaurabh1 made their first contribution in https://github.com/vllm-project/vllm/pull/20413
* @yannqi made their first contribution in https://github.com/vllm-project/vllm/pull/23246
* @jaredoconnell made their first contribution in https://github.com/vllm-project/vllm/pull/23306
* @paulpak58 made their first contribution in https://github.com/vllm-project/vllm/pull/22845
* @zhuangqh made their first contribution in https://github.com/vllm-project/vllm/pull/23309
* @tvalentyn made their first contribution in https://github.com/vllm-project/vllm/pull/23270
* @arjunbreddy22 made their first contribution in https://github.com/vllm-project/vllm/pull/22495
* @philipchung made their first contribution in https://github.com/vllm-project/vllm/pull/17149
* @FoolPlayer made their first contribution in https://github.com/vllm-project/vllm/pull/23241
* @namanlalitnyu made their first contribution in https://github.com/vllm-project/vllm/pull/23375
* @hickeyma made their first contribution in https://github.com/vllm-project/vllm/pull/23353
* @PapaGoose made their first contribution in https://github.com/vllm-project/vllm/pull/23337
* @bppps made their first contribution in https://github.com/vllm-project/vllm/pull/23366
* @AzizCode92 made their first contribution in https://github.com/vllm-project/vllm/pull/23416
* @fengli1702 made their first contribution in https://github.com/vllm-project/vllm/pull/22527
* @FFFfff1FFFfff made their first contribution in https://github.com/vllm-project/vllm/pull/23408
* @ayushsatyam146 made their first contribution in https://github.com/vllm-project/vllm/pull/23171
* @patemotter made their first contribution in https://github.com/vllm-project/vllm/pull/23574
* @Terrencezzj made their first contribution in https://github.com/vllm-project/vllm/pull/23584
* @Copilot made their first contribution in https://github.com/vllm-project/vllm/pull/23385
* @oneraghavan made their first contribution in https://github.com/vllm-project/vllm/pull/23630
* @lordmathis made their first contribution in https://github.com/vllm-project/vllm/pull/23634
* @OYE93 made their first contribution in https://github.com/vllm-project/vllm/pull/23565
* @TianyuLi0 made their first contribution in https://github.com/vllm-project/vllm/pull/23146
* @yuekaizhang made their first contribution in https://github.com/vllm-project/vllm/pull/23623
* @coval3nte made their first contribution in https://github.com/vllm-project/vllm/pull/23054
* @youzhedian made their first contribution in https://github.com/vllm-project/vllm/pull/23648
* @frank-wei made their first contribution in https://github.com/vllm-project/vllm/pull/23613
* @faaany made their first contribution in https://github.com/vllm-project/vllm/pull/22500
* @cndoit18 made their first contribution in https://github.com/vllm-project/vllm/pull/23718
* @rebel-hongseok made their first contribution in https://github.com/vllm-project/vllm/pull/23746
* @Hanchenli made their first contribution in https://github.com/vllm-project/vllm/pull/23713
* @Shrey1306 made their first contribution in https://github.com/vllm-project/vllm/pull/23556
* @Ithanil made their first contribution in https://github.com/vllm-project/vllm/pull/23155
* @killershrimp made their first contribution in https://github.com/vllm-project/vllm/pull/23792
* @crischeng made their first contribution in https://github.com/vllm-project/vllm/pull/23823
* @angelayi made their first contribution in https://github.com/vllm-project/vllm/pull/23349
* @jeanschmidt made their first contribution in https://github.com/vllm-project/vllm/pull/23757
* @He-Jingkai made their first contribution in https://github.com/vllm-project/vllm/pull/23829
* @aditchawdhary made their first contribution in https://github.com/vllm-project/vllm/pull/23899
* @EduardDurech made their first contribution in https://github.com/vllm-project/vllm/pull/23068
* @dubejf made their first contribution in https://github.com/vllm-project/vllm/pull/23944
* @sadeghja1070 made their first contribution in https://github.com/vllm-project/vllm/pull/23971
* @DevonPeroutky made their first contribution in https://github.com/vllm-project/vllm/pull/22769
* @hypdeb made their first contribution in https://github.com/vllm-project/vllm/pull/23802
* @DamonJiang777 made their first contribution in https://github.com/vllm-project/vllm/pull/24028
* @cberge908 made their first contribution in https://github.com/vllm-project/vllm/pull/23104
* @lkm2835 made their first contribution in https://github.com/vllm-project/vllm/pull/23918
* @nathanrchn made their first contribution in https://github.com/vllm-project/vllm/pull/24100
* @co63oc made their first contribution in https://github.com/vllm-project/vllm/pull/24071
* @divyanshsinghvi made their first contribution in https://github.com/vllm-project/vllm/pull/23907
* @biba10 made their first contribution in https://github.com/vllm-project/vllm/pull/23931
* @dongbo910220 made their first contribution in https://github.com/vllm-project/vllm/pull/20388
* @NagyGeorge made their first contribution in https://github.com/vllm-project/vllm/pull/23460
* @wdhongtw made their first contribution in https://github.com/vllm-project/vllm/pull/24214
* @bingchen-mi made their first contribution in https://github.com/vllm-project/vllm/pull/23652
* @anthonsu made their first contribution in https://github.com/vllm-project/vllm/pull/23766
* @whx-sjtu made their first contribution in https://github.com/vllm-project/vllm/pull/23332
* @pratapyash made their first contribution in https://github.com/vllm-project/vllm/pull/24231
* @samanamp made their first contribution in https://github.com/vllm-project/vllm/pull/24266
* @mohankku made their first contribution in https://github.com/vllm-project/vllm/pull/24335
* @ashwin-phadke made their first contribution in https://github.com/vllm-project/vllm/pull/24339
* @bangshengtang made their first contribution in https://github.com/vllm-project/vllm/pull/24265
* @charlotte12l made their first contribution in https://github.com/vllm-project/vllm/pull/23868
* @alhridoy made their first contribution in https://github.com/vllm-project/vllm/pull/24361
* @what-in-the-nim made their first contribution in https://github.com/vllm-project/vllm/pull/24332
* @BraveY made their first contribution in https://github.com/vllm-project/vllm/pull/23928
* @R3hankhan123 made their first contribution in https://github.com/vllm-project/vllm/pull/24034
* @csahithi made their first contribution in https://github.com/vllm-project/vllm/pull/24102
* @Prowindy made their first contribution in https://github.com/vllm-project/vllm/pull/24298
* @micah-wil made their first contribution in https://github.com/vllm-project/vllm/pull/24128
* @pwschuurman made their first contribution in https://github.com/vllm-project/vllm/pull/23845
* @comsky made their first contribution in https://github.com/vllm-project/vllm/pull/24348
* @li-jinpeng made their first contribution in https://github.com/vllm-project/vllm/pull/24519
* @baonudesifeizhai made their first contribution in https://github.com/vllm-project/vllm/pull/23978
* @lacora made their first contribution in https://github.com/vllm-project/vllm/pull/24549
* @RoadToNowhereX made their first contribution in https://github.com/vllm-project/vllm/pull/24217
* @zzhx1 made their first contribution in https://github.com/vllm-project/vllm/pull/23564
* @hjjq made their first contribution in https://github.com/vllm-project/vllm/pull/21078
* @tomasruizt made their first contribution in https://github.com/vllm-project/vllm/pull/24575
* @shengshiqi-google made their first contribution in https://github.com/vllm-project/vllm/pull/24469
* @ca1207 made their first contribution in https://github.com/vllm-project/vllm/pull/23414
* @qandrew made their first contribution in https://github.com/vllm-project/vllm/pull/24127
* @Zazzle516 made their first contribution in https://github.com/vllm-project/vllm/pull/18698
* @RichardoMrMu made their first contribution in https://github.com/vllm-project/vllm/pull/20372

**Full Changelog**: https://github.com/vllm-project/vllm/compare/v0.10.1.1...v0.10.2rc3