Information extraction teams face a recurring choice. Small encoder models are cheap but rigid, and large language models are flexible but expensive per document. Fastino released GLiNER2.5 to narrow ...