Datalab’s Lift is a 9B vision model for structured JSON extraction from PDFs and images, designed to take a PDF or image plus a JSON Schema and return schema-shaped JSON directly in a single pass, rather than converting a document to Markdown first. Lift is best understood as a schema-first document extractor for turning visually complex documents into application-ready fields, distinct from parsers like Docling, MinerU, Marker, and Unstructured, which produce document-shaped intermediate representations. Extractors such as Lift, NuExtract3, LlamaExtract, Reducto Extract, Extend, and Azure Content Understanding output schema-shaped fields. Lift’s approach collapses the parse-then-extract pattern into a single visual extraction pass, reducing pipeline complexity when the goal is field extraction rather than faithful document reconstruction. The competitive map includes open-weight extraction VLMs like NuExtract3, frontier multimodal LLMs, cloud document AI systems, commercial extraction platforms, open-source document parsers, and structured-generation libraries, but not all are direct competitors; Lift is narrower, being the 9B model itself.