🛰️ 服务与 API
OpenAI 兼容 API、router、可观测性。
Push-Based KV Transfer to Improve TTFT and TPOT for Disaggregated Inference in vLLM
本文是『PD 分离服务』路径中 KV 传输优化的进阶篇：应先掌握 P-D disaggregation 基础与默认 NIXL pull connector 的工作流，再读本文的 push 改造。它与双向 KV 传输一文（2026-10-12）互为对照——push 优化单请求首响传输路径但不支持双向传输，两篇合看构成 connector 选型全景。
Adaptive Verification Without a Confidence Head
推测解码路径中自适应验证系列的第二篇:上一篇(2026-08-14 DSpark AV)依赖草稿模型自带的 confidence head,本篇用在线拟合的接受概率估计器去掉该依赖,使任意既有草稿 checkpoint 免重训接入 AV;可继续跟进 AV 在更多 speculator 上的扩展。
Improve TTFT for Multi-Turn Agentic Workloads with Bi-Directional KV Transfer in vLLM
本文是『PD 分离服务』路径中面向多轮/agentic 场景的进阶篇，应在掌握 P-D 分离与默认 NIXL pull connector 后阅读。它与 push connector 一文（2026-10-14）互补：push 优化单请求首响传输路径，本文解决跨轮 KV 复用，且双向能力目前构建在 pull 模式的 NixlConnector 之上。
vLLM Support for NVIDIA Vera Rubin NVL72: 7.8x Throughput over GB200 NVL72
vLLM now runs on NVIDIA Vera Rubin NVL72 with day-0 model support, Rubin-tuned FlashInfer kernels, and locality-aware MoE, delivering 7.8x the per-GPU throughput of GB200 NVL72 on AgentX.
DeepSeek-V4.1-Flash on vLLM: 5x Agentic Throughput Since Day 0
Within three weeks of release, vLLM made DeepSeek-V4.1-Flash 1.9x faster at low concurrency and lifted its throughput 5x on SemiAnalysis AgentX, with SWA bounded replay, CUDA graphs, DeepSeek's new kernels, and vLLM kernel fusions.
Taking vLLM Apart: A Practical Guide to Disaggregated Serving
『PD 分离服务』学习路径的核心实操文章:建议在理解 prefill/decode 差异并读过路由层介绍(如 vLLM Router 一文)之后阅读,再延伸到 Qwen3.8-2.4T PD 容量规划与 bidirectional KV transfer 等专题文章。
Announcing vllm-metal: Concurrent Serving on Apple Silicon
vllm-metal brings vLLM's paged, continuously batched serving stack to Apple Silicon, with flatter TTFT under concurrent agent load, batched MTP, and automatic M5 prefill acceleration.
Scaling Multi-GPU Video Captioning with PyNvVideoCodec and vLLM
How to leverage NVIDIA Hardware Video Decoders to Achieve Multi-GPU Scaling in Video Captioning and Description tasks.
How we trained the fastest DSpark for Kimi-K3 using GB300 NVL72
本文是『推测解码』学习路径中的 DSpark 训练篇:应在掌握推测解码基础与 DFlash/DSpark 算法原理之后阅读,聚焦训练数据(hidden states)如何跨节点规模化生产。它排在两篇自适应验证文章之前,为理解 confidence head 驱动的硬件感知验证调度提供背景与动机。
vime × RL-Kernel × AMD: Bitwise Train–Rollout Consistency on ROCm
vime and RL-Kernel align selected-token logprobs bit for bit across Megatron training and vLLM rollout on AMD Instinct MI300X, with zero mismatches across 200 GRPO steps.
Kimi K3 Performance Optimizations in vLLM: The Road to 2.8× Throughput
Kimi K3 serving optimizations across scheduling, KDA prefix caching, ReplaySSM state recovery, PD disaggregation and state offload, parallelism, MoE, and GPU kernels.
Tiered KV Cache Offloading in vLLM
A host-centric framework for scaling KV cache across host memory, filesystems, object stores, and remote peers — reducing recomputation and increasing serving capacity.
vLLM x AgentX: Optimizing for Real-World Agentic Serving
How vLLM optimizes KV cache management, parallelism, scheduling, and P/D disaggregation for agentic workloads, validated on SemiAnalysis AgentX with up to 130K tokens per GPU-second and a 14.6x-106x serving-cost advantage over Opus 5.
Serving LLMs on Tenstorrent Hardware: Inside the vLLM TT Plugin
Tenstorrent accelerators join vLLM as an out-of-tree platform plugin, driven by mesh-architecture choices: phase-based scheduling, single-process data parallelism on Galaxy, on-device sampling with host fallback, and async decode overlap.
VeRL-Omni v0.2.0: Faster Diffusion RL and Stable Omni Training
A release focused on higher-throughput diffusion rollout, reusable omni adapters, and broader recipe coverage.
Announcing Day-0 Support for NVIDIA Nemotron 3.5 Lightning on vLLM
How vLLM serves NVIDIA Nemotron 3.5 Lightning with OpenAI-compatible APIs, speculative decoding, and BF16/NVFP4 checkpoints across NVIDIA GPUs and edge systems.
vLLM Reaches 25K Total TPS/GPU on Qwen3.5
本文是『PD 分离服务』路径末尾三篇生产规模实战中建议最先读的一篇：它是 Qwen3.8 篇的直接前置（后者开篇引用本文结果），也是 GB200/GB300 NVL72 + 混合 GDN 架构 PD serving 的基础认知入口，可再顺读 GLM-5.2 篇对照 SLA 导向的取舍。
Kimi K3 Is Here: Efficient Day-0 Support on vLLM
vLLM delivers day-0 Kimi K3 serving with hybrid KDA prefix caching, DSpark speculative decoding, production-scale disaggregation, and optimized kernels across NVIDIA and AMD GPUs.
Announcing vLLM AFD Plugin: Disaggregating Attention and FFN for Flexible MoE Serving
『PD 分离服务』路径的横向扩展篇:在掌握按阶段拆分的 PD 分离之后阅读,理解按组件(Attention vs FFN)拆分的另一维度;其异步 connector 本就部署在 PD 分离之上,与 hybrid SSM 篇(同一 PD 框架下的模型适配)互为补充。
From Day 0 to Production SLAs: Serving GLM-5.2 on 24 NVIDIA B300 GPUs with vLLM
本文是『PD 分离服务』学习路径末尾三篇生产规模实战之一，以 SLA 合规而非峰值吞吐为第一目标，与两篇 Qwen 吞吐导向文章形成方法论对照。建议在读完 Qwen3.5（25K TPS）篇建立 NVL72 + 混合架构 PD 基础后阅读，重点体会约束驱动的配置取舍与边界问题定位。
Beyond a Single Model: Building Mixture-of-Models Systems with vLLM Semantic Router
vLLM Semantic Router is expanding from intelligent routing into a system for building, evaluating, and running Mixture-of-Models.
Keeping vLLM Production Quality: A Look Inside CI, Benchmarking, and the Release Process
How vLLM maintains production quality with extensive CI across diverse accelerators, nightly performance benchmark and accuracy evaluation, and a two-week release process.
vLLM x TileRT: Specialized Decode for Latency-Critical Serving
『PD 分离服务』路径的生态应用篇:先理解 PD 分离动机与 vLLM V1 connector 接口,再读本文看 decode 侧可插拔的实际落地;三篇中最易读,适合作为路径收官时对“分离改变推理栈形态”的总结案例。
vime + ROCm: End-to-End RL Post-Training on AMD Instinct™ GPUs
Announcing ROCm support for vime, now running end-to-end on AMD Instinct MI355X GPUs with prebuilt container.
Experience and Lessons Learned from Serving Multi-Stage Qwen3-Omni in vLLM-Omni
How vLLM-Omni serves and optimizes Qwen3-Omni with staged Thinker-Talker-Code2Wav execution, batching, CUDA Graphs, async chunk, async output, replicas, hot-path cleanup, and perf validation.
Micro-Agent: Beat Frontier Models with Collaboration inside Model API
How vLLM Semantic Router turns vllm-sr/auto into a bounded micro-agent runtime for Confidence, Ratings, ReMoM, Fusion, Workflows, and benchmark-shaped collaboration.
Beyond One Model: Fusion in vLLM Semantic Router
How vLLM Semantic Router Fusion runs a panel of models, uses a judge to analyze agreement and gaps, and synthesizes one answer while preserving routing policy, traces, and OpenAI-compatible serving.
Announcing vime: A Simple, Stable, and Efficient RL Framework for LLMs
vime connects slime's training stack with vLLM rollouts to provide a simple, stable, and efficient RL post-training pipeline.
vLLM Semantic Router v0.3 Themis: From Signals to Stateful Production Routing
What vLLM Semantic Router v0.3 Themis adds for production routing: canonical config, inspectable signal-decision-policy flows, safer operations, CLI/dashboard/Kubernetes alignment, and replayable routing behavior.
Announcing Day-0 Support for NVIDIA Nemotron 3 Ultra on vLLM
How to serve NVIDIA Nemotron 3 Ultra with vLLM for long-running agentic reasoning, including BF16 and NVFP4 checkpoints, supported GPU configurations, OpenAI-compatible deployment, and NeMo RL integration.
Fast & Efficient LLM Inference with vLLM: A New Course with DeepLearning.AI
What the DeepLearning.AI vLLM course teaches: optimizing, deploying, and benchmarking LLM inference with LLM Compressor quantization, GuideLLM, KV cache sizing, serving, and memory tradeoffs.
Accelerating vLLM-Omni Inference with AutoRound Quantization
How AutoRound integrates with vLLM-Omni to serve W4A16 quantized multimodal, diffusion, image, and video models with smaller checkpoints, preserved quality, Intel XPU acceleration, and NVIDIA GPU support.
Session-Aware Agentic Routing: Continuity-Aware Model Selection for Long-Horizon LLM Agents
How Session-Aware Agentic Routing in vLLM Semantic Router preserves long-horizon agent continuity with session memory, safe model-switch boundaries, prefix-cache-aware switch pricing, and replayable traces.
vLLM on the DGX Spark: Architecture, Configuration, and Local Evaluation
How to run vLLM on NVIDIA DGX Spark and GB10 systems, including unified memory behavior, NVFP4 Nemotron-3-Super serving, Docker deployment, Prometheus metrics, and local evaluation results.
From Text to Multimodal Routing: Hardening Vision Signals in vLLM Semantic Router
How vLLM Semantic Router hardens multimodal routing by turning visual evidence into trustworthy signals, debugging a Rust/Candle vision-encoder parity issue, and validating image signal correctness for production policy.
Native RL APIs in vLLM
How vLLM native RL APIs standardize weight syncing and asynchronous RL serving with NCCL and CUDA IPC transfer backends, pause mode, and fixes for fragile DPEP and disaggregated rollout deployments.
vLLM x Novita AI: PegaFlow for Production-Grade External KV Cache
How PegaFlow integrates with vLLM as an external KV cache service, using a Rust daemon, CUDA IPC, RDMA, SSD caching, and the external KV connector to improve startup, sharing, throughput, and cache lifecycle.
Announcing VeRL-Omni: Easy, Fast, and Stable RL Training for Diffusion and Omni-Modality Models
How VeRL-Omni extends verl with vLLM-Omni for reinforcement learning post-training of diffusion and multimodal generative models, including efficient rollouts, reward inference, trainers, hardware support, and recipes.
Elastic Expert Parallelism in vLLM
How Elastic Expert Parallelism lets vLLM scale Mixture-of-Experts serving up or down at runtime by changing data-parallel workers, redistributing experts, and coordinating live topology changes without server restarts.
vLLM Tops the Artificial Analysis Leaderboard
How vLLM achieved leading Artificial Analysis results for DeepSeek V3.2, MiniMax-M2.5, and Qwen 3.5 397B using open-source kernel fusion, speculative decoding, Blackwell optimizations, and model-specific serving work.
A First Comprehensive Study of TurboQuant: Accuracy and Performance
A vLLM study comparing TurboQuant KV-cache quantization with BF16 and FP8 across long-context and reasoning workloads, showing where 4-bit variants help, where accuracy drops, and why FP8 remains the default choice.
Serving Agentic Workloads at Scale with vLLM x Mooncake
How vLLM integrates Mooncake Store as a distributed KV cache for agentic workloads, reusing shared prefixes across turns and instances to improve throughput, TTFT, end-to-end latency, and multi-GPU scaling.
Run Highly Efficient Multimodal Agentic AI with NVIDIA Nemotron 3 Nano Omni Using vLLM
How to serve NVIDIA Nemotron 3 Nano Omni with vLLM for multimodal agentic AI, including BF16, FP8, and NVFP4 checkpoints, vision/audio/video inputs, supported GPUs, OpenAI-compatible APIs, and deployment recipes.
Disaggregated Serving for Hybrid SSM Models in vLLM
『PD 分离服务』路径的进阶深化篇:建议先读 vLLM 基础 PD 分离文章(介绍 NIXL connector 的 v0-disagg-prefill)掌握标准流程,再读本文理解 hybrid SSM 模型带来的底层改造;可与 AFD Plugin(按组件拆分)、TileRT(decode 可插拔)两篇生态文对照阅读。
vLLM Korea Meetup 2026 Wrap-Up
What the vLLM Korea Meetup 2026 covered: community growth, vLLM V1 updates, production stack adoption, accelerator integration, vllm-playground, and real-world LLM serving.
Next-Level Inference: Why Your Single-Node vLLM Setup Needs Prefill-Decode Disaggregation
How single-node prefill/decode disaggregation in vLLM uses AMD MORI-IO on an 8-GPU MI300X node to separate prefill and decode, transfer KV cache efficiently, stabilize ITL, and improve goodput.
Run Highly Efficient and Accurate Multi-Agent AI with NVIDIA Nemotron 3 Super Using vLLM
How to serve NVIDIA Nemotron 3 Super with vLLM for multi-agent AI, including BF16, FP8, and NVFP4 checkpoints, 1M-token context, Thinking Budget, MTP, supported GPUs, and OpenAI-compatible deployment.
vLLM Semantic Router v0.2 Athena: ClawOS, Model Refresh, and the System Brain
What vLLM Semantic Router v0.2 Athena adds: refreshed multilingual and multimodal routing models, ONNX and ROCm acceleration, safety and memory signals, long-context handling, and ClawOS orchestration.
Efficiently serve dozens of fine-tuned models with vLLM on Amazon SageMaker AI and Amazon Bedrock
How vLLM serves many fine-tuned MoE and dense models with Multi-LoRA, including fused MoE LoRA kernels, Triton compiler fixes, Split-K and CTA swizzling optimizations, and SageMaker AI and Bedrock tuning.
DeepSeek-V3.2 on GB300: Performance Breakthrough
What DeepSeek-V3.2 and DeepSeek-R1 benchmark results show on NVIDIA GB300 with vLLM, covering NVFP4 quantization, TP and EP deployment, throughput, and reproducible setup details.
GPT-OSS Performance Optimizations on NVIDIA Blackwell: Pushing the Pareto Frontier
How vLLM and NVIDIA optimized GPT-OSS on Blackwell with FlashInfer, torch.compile fusion, FP8 KV cache, async scheduling, stream interval tuning, and deployment recipes that improve throughput and interactivity.
Building Mixture-of-Models on AMD GPUs with vLLM-SR
How vLLM Semantic Router builds a Mixture-of-Models system on AMD MI300X and MI355X GPUs, routing across specialized models with signals, decisions, safety checks, semantic caching, and live MoM deployment.
vLLM Semantic Router v0.1 Iris: The First Major Release
What vLLM Semantic Router v0.1 Iris introduces: signal-decision plugin architecture, model selection, safety filtering, semantic caching, hallucination detection, LoRA-based routing models, and production-ready MoM routing.
Introducing vLLM Playground: A Modern Web Interface for Managing and Interacting with vLLM Servers
How vLLM Playground provides a web UI for starting, configuring, testing, and monitoring vLLM servers across local macOS, Linux GPU or CPU, Kubernetes, and OpenShift environments.
AMD × vLLM Semantic Router: Building the System Intelligence Together
How AMD and vLLM Semantic Router build GPU-accelerated Mixture-of-Models routing with signals, semantic caching, response storage, PII, jailbreak, and hallucination guardrails.
Encoder Disaggregation for Scalable Multimodal Model Serving
How vLLM EPD separates visual encoders from text prefill and decode, covering LMM serving, GPU resource scaling, multimodal interference, and pipelined execution.
Run Highly Efficient and Accurate AI Agents with NVIDIA Nemotron 3 Nano on vLLM
How to serve NVIDIA Nemotron 3 Nano with vLLM for efficient agentic AI, including BF16, FP8, and NVFP4 checkpoints, 1M-token context, hybrid MoE architecture, Thinking Budget, supported GPUs, and OpenAI-compatible deployment.
Token-Level Truth: Real-Time Hallucination Detection for Production LLMs
How HaluGate adds token-level hallucination detection to vLLM Semantic Router by verifying assistant claims against tool outputs and grounding context in real time without LLM-as-judge overhead.
vLLM Router: A High-Performance and Prefill/Decode Aware Load Balancer for Large-scale Serving
『PD 分离服务』学习路径的入门/生态篇:篇幅短,快速建立『P/D 分离在生产中需要专门编排层』的认知;建议先于深度实操文章(如 disaggregated serving guide)阅读,为理解 router/proxy 在 P/D 架构中的角色打基础。
Announcing vLLM-Omni: Easy, Fast, and Cheap Omni-Modality Model Serving
What vLLM-Omni adds to the vLLM ecosystem: omni-modality serving for text, image, video, and audio, diffusion and non-autoregressive generation support, disaggregated stages, OpenAI-compatible APIs, and pipelined execution.
Signal-Decision Driven Architecture: Reshaping Semantic Routing at Scale
How vLLM Semantic Router replaces fixed domain classification with signal-decision architecture, combining multi-dimensional signals, AND/OR decision logic, model selection, and plugin orchestration for production routing.
Docker Model Runner Integrates vLLM for High-Throughput Inferencing
How Docker Model Runner integrates vLLM as an inference backend, letting developers run safetensors models with high-throughput serving, PagedAttention, streaming, and OpenAI-compatible APIs from Docker workflows.
Fast and Affordable LLMs serving on Intel Arc Pro B-Series GPUs with vLLM
How vLLM serves LLMs on Intel Arc Pro B-Series GPUs with MoE optimizations, persistent kernels, multi-GPU scaling, LoRA, speculative decoding, structured outputs, and mixed-precision recipes.
Run Multimodal Reasoning Agents with NVIDIA Nemotron on vLLM
How to serve NVIDIA Nemotron Nano 2 VL with vLLM for multimodal reasoning agents, including video understanding, document intelligence, Efficient Video Sampling, 128K context, and OpenAI-compatible deployment.
Chasing 100% Accuracy: A Deep Dive into Debugging Kimi K2's Tool-Calling on vLLM
How vLLM debugged Kimi K2 tool-calling accuracy, covering chat-template compatibility, add_generation_prompt handling, schema validation failures, benchmark fixes, and tool-use reliability.
From Monolithic to Modular: Scaling Semantic Routing with Extensible LoRA
How vLLM Semantic Router refactors its Rust classification layer with modular model support, Qwen3-Embedding, EmbeddingGemma, LoRA-based multi-task classification, and concurrent routing execution.
No More Retokenization Drift: Returning Token IDs via the OpenAI Compatible API Matters in Agent RL
How vLLM's OpenAI-compatible API can return prompt and response token IDs to prevent retokenization drift in agent reinforcement learning, preserving exact sampled sequences for stable on-policy updates.
SemiAnalysis InferenceMAX: vLLM and NVIDIA Accelerate Blackwell Inference
How vLLM and NVIDIA optimize Blackwell inference for SemiAnalysis InferenceMAX, improving gpt-oss 120B and Llama 3.3 70B throughput with FP4 kernels, scheduling work, and Pareto-frontier benchmarking.
vLLM Semantic Router: Next Phase in LLM inference
How vLLM Semantic Router routes requests by intent, covering semantic classification, smart reasoning-path selection, Rust and Candle execution, and Kubernetes Envoy integration for efficient inference.
Serving Geospatial, Vision, and Beyond: Enabling Multimodal Output Processing in vLLM
How vLLM expands beyond text generation to serve geospatial, vision, and other non-autoregressive models with pooling-model support, TerraTorch integration, raw tensor handling, and flexible IO processors.
Introduction to torch.compile and How It Works with vLLM
How torch.compile works inside vLLM, including TorchDynamo graph capture, TorchInductor code generation, custom compiler passes, graph breaks, model compilation strategy, and performance optimization.
Transformers modeling backend integration in vLLM
How vLLM integrates the Hugging Face Transformers modeling backend to serve more model architectures efficiently, including text and vision-language models through model_impl=\"transformers\".
Llama 4 in vLLM
How vLLM serves Meta Llama 4 Scout and Maverick multimodal MoE models with long-context support, tensor parallel deployment, H100 and H200 guidance, FP8 variants, and performance tips.
PTPC-FP8: Boosting vLLM Performance on AMD ROCm
How PTPC-FP8 quantization improves vLLM performance on AMD ROCm by combining per-token activation scaling and per-channel weight scaling for near-BF16 accuracy with FP8 speed.
vLLM V1: A Major Upgrade to vLLM's Core Architecture
What changed in vLLM V1: a re-architected engine with a simpler scheduler, near-zero-overhead prefix caching, cleaner tensor parallelism, multiprocessing API server, and default optimizations for higher-throughput serving.
Introducing vLLM Inference Provider in Llama Stack
How Llama Stack integrates vLLM as an inference provider through remote and inline providers, enabling OpenAI-compatible vLLM serving for local and Kubernetes generative AI application deployments.
High Performance and Easy Deployment of vLLM in K8S with vLLM production-stack
What vLLM production-stack adds for Kubernetes serving: prefix-aware routing, LMCache-backed KV cache sharing, autoscaling, observability, fault tolerance, and cluster deployment with higher throughput and lower latency.
vLLM 2024 Retrospective and 2025 Vision
A vLLM 2024 retrospective and 2025 roadmap covering community growth, model and hardware support, production adoption, office hours, ecosystem partnerships, and the path toward universal open-source serving.
vLLM v0.6.0: 2.7x Throughput Improvement and 5x Latency Reduction
What changed in vLLM v0.6.0 to improve throughput and latency, including API server isolation, reduced CPU overhead, multi-step scheduling, async execution, and benchmarks against earlier vLLM versions.
vLLM’s Open Governance and Performance Roadmap
vLLM's open governance and performance roadmap, covering LF AI and Data incubation, public benchmarks, optimized kernels, async scheduling, API frontend overhead, torch.compile, disaggregated prefill, and community research.
Notes on vLLM v.s. DeepSpeed-FastGen
A performance comparison of vLLM and DeepSpeed-FastGen, explaining when Dynamic SplitFuse helps, where vLLM is faster, and how memory allocation, output length, and workload shape affect throughput.
vLLM: Easy, Fast, and Cheap LLM Serving with PagedAttention
What the original vLLM launch announced: PagedAttention for KV cache management, up to 24x throughput over Hugging Face Transformers, and lower-cost high-throughput LLM serving.