Cloudflare Releases Clairic: Single-Pass Document Extraction Model Claims Faster and Cheaper Performance Than LLMs

Cloudflare released an AI model named Clairic on October 20, 2026, which the company says pulls data from documents faster and more cheaply than large language models (LLMs). Clairic is currently available in preview on Workers AI, with open weights expected to follow soon.

Unlike language models that generate text token by token, Clairic uses a 'single-pass' mechanism to answer all fields. Users specify desired fields (such as order number or invoice total), and the model returns matching values along with a confidence score, returning 'null' if a value is missing. Beyond extraction, it can handle simple calculations, estimates, and forecasts.

Cloudflare states that Clairic is designed to run as one piece of a larger agent chain. For example, when processing a complaint email, Clairic extracts key information and estimates repair costs, then a decision model called Clef applies company rules to choose between a refund and a repair, and finally GLM-5.3 writes the reply. According to Cloudflare, this chain takes 19.3 seconds, whereas using GLM-5.3 alone for all three steps requires 61.7 seconds and costs almost twice as much. However, the model cannot yet reliably handle longer chains of calculations, requiring developers to run those in their own code.