OpenRouter Launches Structured Data Extraction API with Multimodal Support

On November 10, 2026, OpenRouter announced the launch of the Extractions API, a new endpoint for pulling typed fields out of documents and images. Callers define the fields they need and receive the extracted values together with position spans and confidence scores, so downstream systems can consume structured output without parsing free-form text or doing their own vision inference. The feature was described as supported by three models at launch, and it can run on platforms including Cloudflare, broadening where the workload is executed. As an example of the capability, OpenRouter noted that on Cloudflare's Clairic the same request reads both text and images and returns numbers, dates, lists, estimates and forecasts, positioning the API as a way to turn heterogeneous documents into typed records. The launch announcement did not detail pricing, rate limits or a wider rollout schedule, so the current status of the offering is limited to what was described on that date.