Cloudflare introduced Clairic as a preview model on Workers AI. The company describes it as an extraction model that can return values from unstructured inputs, including extracted fields, calculations, estimates, and forecasts.
A narrower model for structured tasks
- The announcement positions Clairic for fields such as names, dates, totals, and lists taken from input documents.
- Cloudflare says numeric answers include an uncertainty range and extracted spans can include offsets and confidence.
- The model is hosted as a Workers AI preview; the company says it intends to release open weights later.
A structured result is not automatically correct. Validate extracted values against the source, handle missing or ambiguous fields, and use a model suited to the cost and error tolerance of the task.
Preview status and performance claims can change. Check the current model documentation and evaluate representative documents before relying on the output.
