# vLLM v0.11.0 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.11.0
- **Released:** Oct 02, 2025
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.11.0/

## Executive Summary (TL;DR)
vLLM v0.11.0. Source highlight: This release completes the removal of V0 engine. V0 engine code including AsyncLLMEngine, LLMEngine, MQLLMEngine, all attention backends, and related components have been removed. **V1 is the only engine in the codebase now.**

## Breaking Changes & Deprecations
- [Spec Decode][CI] Add e2e test for `examples/spec_decode.py` and prevent breaking Acceptance Length by @ekagra-ranjan in https://github.com/vllm-project/vllm/pull/24531

## New Features
- Features: Qwen3-VL text-only mode (#26000), EVS video token pruning (#22980), Mamba2 TP+quantization (#24593), MRoPE + YaRN (#25384), Whisper on XPU (#25123), LongCat-Flash-Chat tool calling (#24083).
- V1 features: Prompt embeddings (#24278), sharded state loading (#25308), FlexAttention sliding window (#24089), LLM.apply_model (#18465).
- [CI] Add ci_envs for convenient local testing
- [Misc][gpt-oss] Add gpt-oss label to PRs that mention harmony or related to builtin tool call
- [Multi Modal] Add FA3 in VIT
- [CI] Trigger BC Linter when labels are added/removed

## Improvements
- This releases turns on **FULL_AND_PIECEWISE as the CUDA graph mode default**. This should provide better out of the box performance for most models, particularly fine-grained MoEs, while preserving compatibility with existing models supporting only PIECEWISE mode.
- Performance: GLM-4.1V 916ms TTFT reduction via fused RMSNorm (#24733), GLM-4 MoE SharedFusedMoE optimization (#24849), Qwen2.5-VL CUDA sync removal (#24741), Qwen3-VL Triton MRoPE kernel (#25055), FP8 checkpoints for Qwen3-Next (#25079).
- Performance: FlashInfer RoPE 2x speedup (#21126), fused Q/K RoPE 11% improvement (#24511, #25005), 8x spec decode overhead reduction (#24986), FlashInfer spec decode with 1.14x speedup (#25196), model info caching (#23558), inputs_embeds copy avoidance (#25739).
- NVIDIA: FP8 FlashInfer MLA decode (#24705), BF16 fused MoE for Hopper/Blackwell expert parallel (#25503).
- DeepGEMM: Enabled by default (#24462), 5.5% throughput improvement (#24783).
- New architectures: RISC-V 64-bit (#22112), ARM non-x86 CPU (#25166), ARM 4-bit fused MoE (#23809).

## Bug Fixes
- Intel XPU: MoE DP accuracy fix (#25465).
- Fix implementation divergence for BLOOM models between vLLM and HuggingFace when using prompt embeds
- [Bugfix] Fix MRoPE dispatch on CPU
- [BugFix] Fix Qwen3-Next PP
- [CI] Fix flaky test  v1/worker/test_gpu_model_runner.py::test_kv_cache_stride_order
- [Bugfix] Fix BNB name match

## Upgrade Command
```bash
python -m pip install 'vllm==0.11.0'
```