Schematron V2 Small

Schematron V2 Small is Inference.net's lightweight specialized extraction model, built specifically to convert raw, messy web pages into neat, organized data. Instead of writing conversational replies, it reads long source text and directly extracts requested information—such as pulling a product's name, price, and location out of an unstructured listing—into a strictly defined format without needing conversational instructions. Because it is compact and capable of processing extensive pages in one pass, it handles information extraction smoothly and rapidly.

It is well suited for developers and data teams building automated web scraping pipelines that require reliable, schema-conforming structured output at a low running cost. Within the Schematron family, it delivers the highest extraction quality, matching the accuracy of much larger models on intricate schemas and long documents while keeping expenses modest. Since it is purpose-built to extract information according to a target layout, it is designed for structured data transformation rather than creative writing or general dialogue.