# vLLM v0.21.0 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.21.0
- **Released:** May 15, 2026
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.21.0/

## Executive Summary (TL;DR)
vLLM v0.21.0. Source highlight: This release features 367 commits from 202 contributors (49 new)!

## Breaking Changes & Deprecations
- C++20 build requirement**: vLLM now requires a C++20-compatible compiler for compatibility with PyTorch (#40380). This is a **breaking build change**.
- Breaking**: C++20 required for PyTorch compatibility (#40380).
- Breaking**: Transformers v4 deprecated (#40389).

## Improvements
- **NVIDIA Blackwell**: TOKENSPEED_MLA backend for DSR1/Kimi-K25 (#41778), faster per-token FP8 group quant packed kernel (#41326), FP8 on NVIDIA Thor/SM110 (#39712), CUTLASS scaled mm for non-compatible sizes (#41868).
- **Performance**: FlashInfer top-k/top-p sampler enabled by default (#40376), FP8 FlashInfer attention for ViT (#38065), TurboQuant shared dequant buffers (#40941), `AllPool.forward` 51% faster (#41163), GPU<->CPU sync elimination in pooling (#41433) and attention (#41434), numpy zero-copy embedding serialization (#41681), multimodal processor skip for text-only (#41246), FlashInfer FP8 async TP fusion (#39505), NVFP4 all-gather GEMM fusion for AsyncTP (#41882), re-enable allreduce+RMS fusion for
- **CPU**: FP8 attention for AMX/AVX-512 (#39445), FP8 W8A16 linear (#41186), FP8 W8A16 MoE (#41314), DNNL AVX2 W8A8 Int8 (#41318), Gated DeltaNet Attention for Qwen 3.5/3.6 (#41025), RISC-V OMP thread auto-binding (#40569).
- **Intel XPU**: Top-k/top-p sample kernel (#39285), out-of-place all-reduce (#41808), LoRA support (#38206).
- **IBM Power**: VSX attention backend (#40451).
- **FlexAttention**: Re-enabled for batch invariant mode (#40842).

## Bug Fixes
- DeepSeek V4: AMD/ROCm support (#40871), pipeline parallelism (#41694), `max` reasoning effort (#40982), disaggregated serving fixes (#41957).
- Gemma3/Gemma4: `hidden_act` variant support (#40588), pipeline parallelism fix (#40786), MoE fixes (#41206, #41574, #41401), tool parser crash fix (#41991, #42188).
- Model Runner V2: Rejection sampling acceptance rate fix (#40651), skip metadata rebuild before draft prefill (#40410), rebuild metadata between draft decode steps (#41162), Qwen3.5/Mamba hybrid support (#35520).
- Ray: RayExecutorV2 enabled by default (#41421), actor name collision fix for DP > 1 (#40398).
- **AMD ROCm**: ROCm 7.2.2 (#41386), DBO (Dynamic Batch Optimization) (#34726), AITER Fused Allreduce+RMSNorm (#37646), Fused Shared Expert (FSE) for Qwen3-Next (#39280), DeepSeek V3.2 TP4 AITER MLA (#41835), GDN linear attention fusion (#40711), eliminate redundant MoE buffer copies in AITER (#41713), CPU offloading support (#40549), DeepEP API update (#39721), cap Triton paged attention block size to fix shared memory OOM (#38502).
- LoRA: Initial expert parallel (EP) support (#40867), Qwen3.5 LoRA fusion fix (#37912).

## Upgrade Command
```bash
python -m pip install 'vllm==0.21.0'
```