Cloudflare has launched Clairic, a model designed to extract specific data from documents more quickly and cheaply than large language models. Available in preview on Workers AI, the tool accepts user-defined fields such as order numbers or invoice totals and returns matching values with a confidence score. Unlike generative models that produce text token by token, Clairic answers every requested field in a single pass. If a value does not exist, the system returns null. Beyond simple extraction, the model can perform basic calculations, estimates, and forecasts.
The technology is intended to run in chains alongside LLMs and other system models like Jev. In a customer complaint scenario, Clairic pulls the order number, amount, and damage details, then estimates repair costs. A decision model called Clef applies company rules to choose between a refund and a repair, while GLM-5.3 writes the final reply. Cloudflare states this chain takes 19.3 seconds, whereas GLM-5.3 alone requires 61.7 seconds for the same steps and costs nearly twice as much. The company notes Clairic cannot yet reliably handle longer chains of calculations, so developers should run those in their own code.
- Open weights are set to follow soon
- System runs in a single pass per field
- Missing values return null instead of hallucinating
