Datalab releases lift, a 9B open-weights vision model for PDFs

MarkTechPost reported that Datalab has released lift, a 9B open-weights vision model for extracting structured data from PDFs. The model is designed to output schema-matching JSON and is reported to reach 90.2% field accuracy. The release matters for teams building document-heavy AI workflows because PDF extraction remains a common failure point in enterprise automation, compliance, research, and financial operations.

Key details: Published June 23, 2026 at 19:35 UTC, Datalab released lift as a 9B open-weights vision model, The model extracts schema-matching JSON from PDFs, MarkTechPost reports 90.2% field accuracy.

Why it matters: Open document models can lower the cost of building specialized enterprise extraction workflows, especially where proprietary OCR systems are expensive or hard to adapt.