# vLLM v0.19.0 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.19.0
- **Released:** Apr 03, 2026
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.19.0/

## Executive Summary (TL;DR)
vLLM v0.19.0. Source highlight: This release features 448 commits from 197 contributors (54 new)!

## New Features
- Features: Temporal compression for Nemotron-3-VL videos (#36808), NemotronH Puzzle + MTP (#37803), torch.compile for InternVL vision encoder (#38049), multiple embedding types in single call (#35829).
- **Features**: Limit thinking tokens (hard limit) (#20859), multiple embedding types in single call (#35829), numpy array embeddings for multimodal (#38119), `--lora-target-modules` (#34984), `-sc` shorthand for `--speculative-config` (#38380).
- Add `VLLM_MAX_N_SEQUENCES` environment variable to enforce sequence limits (#37952).
- flashinfer-cubin added as default CUDA dependency (#37233).

## Improvements
- **Zero-bubble async scheduling + speculative decoding**: Async scheduling now supports speculative decoding with zero-bubble overlap, significantly improving throughput (#32951).
- Performance: GLM-4.xv ViT optimization (#37779).
- B300/GB300 (SM 10.3): Allreduce fusion enabled by default (#37755), tuned all-reduce communicator (#37756).
- FlashInfer sparse MLA as default for FP8 KV cache (#37252).
- Tuned prefill configs for FP8 FA3 (#36265), tuned Triton MoE config for Qwen3.5 on H200 with 9.9% E2E improvement (#37340), H800 MoE configs (#31201).
- GPT-OSS: Router GEMM kernel (#37205), eliminate padding with FlashInfer MXFP4/MXFP8 MoE (#30647), reduce redundant SparseMatrix creation (#37683).

## Bug Fixes
- **Transformers v5 compatibility**: Broad compatibility fixes across many models for HuggingFace Transformers v5 (#37681, #38127, #38090, #38247, #38410).
- Speculative decoding: Eagle3 for Pixtral (#37182), EagleMistralLarge3 fix (#37232).
- LoRA expansion: H2OVL tower/connector LoRA (#31696), `--lora-target-modules` to restrict LoRA to specific modules (#34984), `language_model_only` respected (#37375), Mistral3 fix (#36928), Qwen3.5 fix (#36976), out-of-tree ops replacement (#37181).
- Model fixes: NemotronH MTP + Chunked Prefill (#35447), Qwen3-VL video timestamps (#37439), Qwen3.5 GDN quantized models (#37448), Qwen3Next A_log FP32 (#37810), JAIS ALiBi (#37820), RoBERTa CUDA graph position IDs (#37873), AudioFlamingo3/MusicFlamingo (#37643), Music Flamingo loading (#35535), bge-m3 task selection (#37632), Nemotron Parse loading (#37407), GLM OCR patch merger (#37962), PaddleOCR checkpoint compat (#38232), DeepSeek v3.2 params (#33703), MiniMax NVFP4 weight loading (#37214), 
- **Performance**: Optimize top-k in Triton sampler (#37225), optimize token_embed for pooling models with 1% improvement (#37347), fix slow hasattr in CUDAGraphWrapper (#37425), NFS prefetch auto-enabled with RAM guard (#37673), pybase64 replacement (#37290), optimize swap_states for hybrid models (#34733).
- **Bugfixes**: Fix gibberish from FP8 MLA KV scale inconsistency (#37054), Mamba state corruption (#37728), deadlock with pause/resume (#37024), FlashInfer MNNVL socket collisions (#36674), multimodal prefix cache key collisions (#36708), DP coordinator ZMQ TOCTOU (#37452), CUDA graph memory double-counting (#37426), pooling non-determinism (#37775), AllReduce Fusion shutdown crash (#36955), FlashInfer allreduce workspace (#37461), async spec decoding with hybrid models (#38556), MLA sparse index

## Upgrade Command
```bash
python -m pip install 'vllm==0.19.0'
```