Tool · 2026-07-11

vLLM v0.25.0

vLLM v0.25.0 ships 558 commits from 232 contributors. Model Runner V2 becomes the default for all dense models, building on the quantized-model support introduced in the previous release.

vLLM Releases2026-07-11About 2 min readWorth tracking

Marked important, but published more than 30 days ago, so it moved down to this band.

Inference & deployment
On this page13 sections

The full write-up and the editors' notes are in Chinese; the summary above is from the original announcement.

Release line

This signal has a follow-up

#

You are reading an earlier release on this line. 4 more releases follow; the latest is “v0.32.0”.

  1. 2026-06-30vLLM v0.24.0
  2. 2026-07-11vLLM v0.25.0Current
  3. 2026-07-25vLLM v0.26.0
  4. 2026-08-10vLLM v0.27.0
  5. 2026-09-22v0.30.0
  6. 2026-10-18v0.32.0Latest

承接 v0.24.0 稳定版：Model Runner V2 从量化模型扩到所有稠密模型，推理执行栈完成代际切换。

Record

The signal

#

vLLM v0.25.0 是推理执行栈切换的收官版本：Model Runner V2（MRv2）正式成为所有稠密模型的默认执行引擎。这次发布汇集了 558 个提交、232 位贡献者（其中 64 位是新贡献者），延续了上一个版本（v0.24.0）打下的量化模型支持基础。

对使用者来说，这是一次「默认值变更」级别的升级：过去需要显式开启的新执行引擎，现在开箱即用。执行引擎决定了批处理调度、KV 缓存管理和算子路径，直接影响吞吐与延迟表现，因此升级前后值得用自己的真实负载做一轮基准对比，而不是只看官方数字。

结合同期 Hugging Face 宣布 Transformers 后端在 vLLM 中达到原生速度，可以看到开源推理生态正在收敛：vLLM 负责极致的服务优化，模型实现层则越来越多地交给标准化的建模库。对自建推理服务的团队而言，vLLM 的大版本节奏（v0.24 → v0.25 两个月内完成执行栈切换）本身就是一个值得跟踪的信号。

Why now

Why it matters

#

推理引擎的默认执行栈切换会直接改变吞吐、延迟和 GPU 成本曲线。vLLM 是目前自建大模型服务最主流的开源选择，它的大版本升级几乎影响所有自托管推理团队。

Background

Technical context

#

Model Runner V2 是 vLLM 重写的模型执行层，负责调度、KV 缓存与算子路径。v0.24 先以量化模型为试点，v0.25 将其推广为全部稠密模型的默认值——这是典型的「灰度到默认」的执行栈迁移路径。

At your level

Have AI explain this signal at your level

Pick your level of experience and AI writes an explanation for it on the spot.

Audience

Who should care

#
推理平台与服务工程师关注 GPU 成本的 MLOps 负责人自托管开源模型的技术团队
Areas it may affect
自托管模型推理服务GPU 成本与容量规划开源推理生态标准化

Next

Learning path

#
  1. 先理解推理引擎在服务栈中的位置：模型实现、执行引擎、服务编排各自负责什么
  2. 用官方文档跑通 vLLM 的基本部署，观察吞吐与显存占用
  3. 在自己的真实负载上对比 v0.24 与 v0.25 的表现，验证默认值变更的影响

Getting started

Have AI draft a learning path

From the related knowledge and skills on this page, AI drafts a path from the basics to applying it.

Network

Where it sits in the network

#

How this signal connects to neighbouring technologies, skills and background knowledge. Follow any link to keep exploring.

vLLM v0.25.0
Signals · 2
Skills · 3
Knowledge · 5
View in the network →

Comparison

Compare with another technology

Pick another published technology and AI writes a comparison of what they share, how they differ and when to use each.

Further thought

Follow-up questions

#
  • MRv2 对 MoE 与多模态模型的支持进度如何？
  • 你的负载在 v0.25 上的吞吐/延迟变化是多少？
  • Transformers 后端与原生实现在你的目标模型上还有差距吗？

Publication record

Featured in

#

The briefings that carried this signal, newest first; the last row is the weekly review of its week. The tag is the section it ran in.

  1. 技术简报 - 2026-07-15

    Immediate attention

  2. 技术简报 - 2026-07-14

    Immediate attention

  3. 技术简报 - 2026-07-13

    Immediate attention

  4. Weekly review · Jul 6 – Jul 12

    Worth tracking

Sources

vLLM Releases

vLLM ProjectOpen-source community2026-07-11
Open original sourcehttps://github.com/vllm-project/vllm/releases/tag/v0.25.0