Understand this story
UC Berkeley Sky Computing Lab built and open-sourced the vector index LEANN, which does not store full embedding vectors but instead recomputes them on demand at query time. According to a MarkTechPost report, repository benchmarks show that the index for 60 million Wikipedia chunks dropped from 201 GB to 6 GB; in the paper, on a 76 GB corpus, the index was compressed to 4 GB, and downstream accuracy on NQ was 25.5%, on par with HNSW, but retrieval took 2.48 seconds versus 0.05 seconds for HNSW. The work won the MLSys 2026 Best Research Paper award, and the code is released under the MIT license; it can be installed via uv pip install leann, and it provides documentation, ready-made applications such as Chrome history and WeChat, and an MCP retrieval service for Claude Code.
AI-generated from public reports
Latest development · Report
UC Berkeley Open-Sources LEANN Vector Index: 60M-Chunk RAG Index Reduced from 201 GB to 6 GBDevelopments
1 confirmed developments
UC Berkeley Sky Computing Lab open-sources vector index LEANN
Source:MarkTechPost
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 28
MarkTechPost
UC Berkeley Open-Sources LEANN Vector Index: 60M-Chunk RAG Index Reduced from 201 GB to 6 GB
UC Berkeley Sky Computing Lab has open-sourced the LEANN vector index, which does not store full embedding vectors but instead recomputes them on demand at query time. Repository benchmarks show that the index for 60 million Wikipedia chunks drops from 201 GB to 6 GB. In the paper, the index is compressed to 4 GB on a 76 GB corpus, and the NQ downstream accuracy of 25.5% matches HNSW, but retrieval takes 2.48 s, compared with 0.05 s for HNSW. The work received the MLSys 2026 Best Research Paper award, and the code is released under the MIT license. It can be installed via uv pip install leann, and it provides documentation, ready-made applications such as Chrome history and WeChat, and an MCP retrieval service for Claude Code.