Extract

scaledown/extract

Extract is ScaleDown's entity extraction model (1M-token context) where you describe the fields you want in plain English instead of fixed entity types. It returns matching spans with confidence scores and character offsets, and it does not support tool calling or enforced JSON output, so it suits pulling structured facts out of long documents.

Lanzamiento
13 oct 2026
Entrada
TextImageFile
Salida
Text
Compatibilidad
RazonamientoLlamada a herramientasSalidas estructuradas
Extract: precios, contexto y benchmarks | SurfMind