vLLM's Hybrid HiSparse keeps decoding past HBM limits: 19-25 vs 5-6 concurrent 1M-context requests
vllm_project · x · 2026-09-09
The vLLM project (built by RedHat AI, Prime Intellect and community) introduced Hybrid HiSparse: since Sparse MLA only attends to top-K tokens, KV no longer needs to fully reside on GPU, so a request keeps decoding after its KV outgrows HBM instead of being preempted.
Under memory pressure, coldest KV pages are offloaded to host memory while a small hot buffer serves the indexer. Key details: hot pages are ordinary KV blocks from the same pool (Hybrid Memory Allocator); a single fused kernel handles resident, hot and missing rows and is CUDA-graph capturable; prefix caching, OffloadingConnector, P/D imports and MTP all keep working.
Demo: one 8× H200 node, GLM 5.3, full 1M context, concurrency 32 — KV offloading kept 5-6 requests running vs 19-25 with Hybrid HiSparse.
Related event: vLLM Adds Hybrid HiSparse, Quadrupling Long-Context Concurrency(2 posts)→
More from Infra
- MiniMax H3 runs fully on a Galaxy S26: 5-second video with audio in ~7 minutes — BidFunny4862 · 2026-10-27
- Unconventional AI acquires inference chip firm NeuReality in 1,000x efficiency bid — NaveenGRao · 2026-10-27
- Musk: robot RAM demand will dwarf everything else, and 100% of superintelligence compute goes to deep space — elonmusk · 2026-10-27
- RTX 5090 hits $7K as PC gaming community goes into 'denial' over AI boom — PAstynome · 2026-10-27
- Wan2GP lands on Pinokio: 1-click AI video on 6GB GPUs — cocktailpeanut · 2026-10-27
- AWS on agent prompt caching: explicit + implicit patterns to slash token costs — AWS ML Blog · 2026-10-27