Cloudflare has introduced Clairic, an AI model designed for document data extraction. The tool is currently available in preview via the Workers AI platform, with plans to release open weights in the near future.

Unlike traditional large language models that generate text token by token, Clairic processes documents in a single pass. Users define specific fields, such as invoice totals or order numbers, and the model returns the corresponding values along with a confidence score. If data is absent, it returns a null value.

Beyond simple extraction, the model performs basic calculations, estimates, and forecasting. Cloudflare intends for Clairic to function within automated chains alongside other specialized models like Jev, Clef, and GLM-5.3 to handle complex workflows.

In a test scenario involving a complaint email, the chain utilizing Clairic completed the process in 19.3 seconds. In comparison, using only the GLM-5.3 model for the same three-step task required 61.7 seconds and incurred nearly double the cost.

The company notes that Clairic currently lacks the capability to reliably manage extended chains of complex calculations. Developers are advised to handle such operations within their own application code for the time being.