inference-net/Schematron-3B

inference-net/Schematron-3B is a 3.2 billion parameter model from the Schematron series, purpose-trained by Inference.net for converting noisy HTML into clean, typed JSON that conforms to a custom schema. This model excels at schema-first extraction, ensuring 100% schema-conformant JSON outputs from lengthy, noisy HTML up to a 128K token context window. It is primarily designed for web scraping, data ingestion, and transforming arbitrary web pages into structured records.