--- title: vLLM × HPC-Ops: High-Performance Attention and MoE... description: Tencent Hunyuan's HPC-Ops operator library has contributed two production-hardened kernels to vLLM's main branch as first-class backends. The Attention backend... canonical: https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: vLLM × HPC-Ops: High-Performance Attention and MoE Backends from Tencent Hunyuan | daily.dev og:description: Tencent Hunyuan's HPC-Ops operator library has contributed two production-hardened kernels to vLLM's main branch as first-class backends. The Attention backend... og:url: https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s og:image: https://api.daily.dev/og/posts/iVteBw42s.png og:image:alt: vLLM × HPC-Ops: High-Performance Attention and MoE Backends from Tencent Hunyuan og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # vLLM × HPC-Ops: High-Performance Attention and MoE Backends from Tencent Hunyuan **[vLLM](https://daily.dev/sources/vllm)** · 15 min read · 2 upvotes · 0 comments ## Summary Tencent Hunyuan's HPC-Ops operator library has contributed two production-hardened kernels to vLLM's main branch as first-class backends. The Attention backend replaces fixed split-KV scheduling with a per-step load-balanced dynamic scheduler plus a fused RoPE+QK-Norm+KV-write prologue, achieving up to 2.95× speedup over static scheduling on mixed-length decode batches and 2.25× on average over FlashInfer and FlashAttention. The MoE backend fuses the entire MoE pipeline (routing, Gate-Up GEMM, activation/quantization, Down GEMM, top-k reduction) into a compact FP8 execution path using occupancy-first design and PDL-chained stages, delivering 1.59× speedup at TP8/EP1 and 1.21× at TP1/EP8 over Triton and CUTLASS. End-to-end on the Hy3 295B MoE model across 8× H20 GPUs, the combined backends cut TTFT by ~24% and TPOT by ~17% versus vLLM defaults. Both backends integrate via standard vLLM backend interfaces with no source modifications required, and are optimized for NVIDIA Hopper architecture (especially H20). ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [vLLM Triton Attention Backend Deep Dive](https://daily.dev/posts/vllm-triton-attention-backend-deep-dive-kx2nzacro) · vLLM · 0 upvotes · 0 comments - [Beyond Porting: How vLLM Orchestrates High-Performance Inference on AMD ROCm](https://daily.dev/posts/beyond-porting-how-vllm-orchestrates-high-performance-inference-on-amd-rocm-9sjfha0v1) · vLLM · 2 upvotes · 0 comments - [The State of FP8 KV-Cache and Attention Quantization in vLLM](https://daily.dev/posts/the-state-of-fp8-kv-cache-and-attention-quantization-in-vllm-ovwjvspr4) · vLLM · 1 upvotes · 0 comments - [Boosting MoE Training Throughput with Advanced Fusion Kernels](https://daily.dev/posts/boosting-moe-training-throughput-with-advanced-fusion-kernels-drffsfppe) · NVIDIA Developer · 1 upvotes · 0 comments --- Tags: [#ai-inference](https://daily.dev/tags/ai-inference), [#vllm](https://daily.dev/tags/vllm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts) [View this post on daily.dev](https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"vLLM × HPC-Ops: High-Performance Attention and MoE Backends from Tencent Hunyuan","url":"https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s"},"datePublished":"2026-07-06T03:55:12.502Z","dateModified":"2026-07-06T03:55:41.146Z","description":"Tencent Hunyuan's HPC-Ops operator library has contributed two production-hardened kernels to vLLM's main branch as first-class backends. The Attention backend...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/79f7e259ad68f362b4ac838301d7fe2d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/79f7e259ad68f362b4ac838301d7fe2d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"vLLM","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"vLLM","logo":"https://media.daily.dev/image/upload/s--hTxEuls9--/f_auto/v1744613054/logos/vllm","url":"https://daily.dev/sources/vllm"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/vllm-hpc-ops-high-performance-attention-and-moe-backends-from-tencent-hunyuan-ivtebw42s","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-inference,vllm,mixture-of-experts","timeRequired":"PT15M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"vLLM","item":"https://daily.dev/sources/vllm"},{"@type":"ListItem","position":3,"name":"vLLM × HPC-Ops: High-Performance Attention and MoE Backends from Tencent Hunyuan"}]} ```