Open source · 28 October 2026
Available nowUC Berkeley releases LEANN index cutting 60M‑chunk RAG size from 201 GB to 6 GB
LEANN stores a pruned graph and recomputes embeddings at query time, matching HNSW 90% recall using 4 GB instead of 188 GB, with seconds‑level search. It reduces a 60 M‑chunk RAG index from 201 GB to 6 GB.
Why it matters
LEANN reduces index size to 6 GB for 60 M chunks; plan storage‑efficient RAG features, but account for slower query latency.
For product managers
For your product
What does this mean for your product?
Try it on your product More from this issueAll AI news from 28 Oct →
Get this every morning, written for your product.
Deltaday reads the whole AI news cycle and tells you what changed, whether it's real yet, and what to do about it.