MLPerf Storage v3.0 lands with 144 results, adding KV cache and vector DB tests
TheKanter · x · 2026-09-02
MLCommons released MLPerf Storage v3.0 results: 144 results from 19 organizations, 11 first-time submitters. The suite now covers four workloads — training, checkpointing, vector database, and KV cache — with checkpointing tested from 8B up to 1250B parameters.
Key additions:
- A KV Cache test measuring storage performance for LLM inference cache read/write;
- A Vector DB test for high-dimensional indexing and query workloads;
- S3 object storage API support alongside POSIX;
- New efficiency metrics: rack-unit density and provisioned power.
Working group co-chair Brian Belgodere said decomposing monolithic AI systems into storage patterns like checkpointing, KV caching, and vector DBs gives stakeholders a much clearer view of AI storage needs.
More from Infra
- SGLang Summit sponsors grow from 11 to 36 in one week; Lip-Bu Tan, Lilian Weng to speak — BanghuaZ · 2026-10-18
- Yacine jokes his datacenter KPI is gigawatt-hours: get paged if not burning enough power — yacineMTB · 2026-10-18
- 1.68x inference speedup drops, directly accelerating GRPO post-training — yacineMTB · 2026-10-18
- OpenRouter is only a fraction of the total addressable market, poster argues — sloppenheimer · 2026-10-18
- Developer builds change alerts for LLM prices and retirements across Bedrock, Azure and Vertex — jsandhu0780 · 2026-10-18
- 12.3M tokens in 27 minutes for $0.14: agent session with 99.6% cache hit — teortaxesTex · 2026-10-18