Schematron V2 Turbo

Schematron V2 Turbo is Inference.net's lightweight, speed-oriented model designed to turn messy web pages into clean, organized data. Rather than engaging in general conversation, it reads long, cluttered markup and directly pulls out target details—such as extracting a product's title, pricing, technical specifications, and tags from an online listing—into a strictly conforming structured format. Because it is compact and capable of reviewing lengthy documents in a single pass, it handles routine web-harvesting workloads rapidly while keeping operating expenses low.

It is built for developers and data teams running high-volume web scraping or ingestion pipelines where maximum throughput and economical processing are top priorities. Within the Schematron family, it processes pages significantly faster than Schematron V2 Small and exceeds the extraction quality of the earlier Schematron 3B, though V2 Small remains the preferred choice when handling the most complex schema structures. Notably, it does not accept conventional conversational prompts or directions, meaning all extraction instructions and field expectations must be communicated strictly through the defined target schema.