Cloudflare's Clairic pulls data from documents faster and cheaper than LLMs (the-decoder.com)

Cloudflare has launched Clairic, an innovative AI model that extracts data from documents more quickly and affordably than traditional large language models (LLMs). Currently in preview on Workers AI, Clairic allows users to specify fields, such as order numbers or invoice totals, returning the relevant information in a single pass along with a confidence score. Unlike LLMs, Clairic does not generate text incrementally; instead, it efficiently processes requests, even executing simple calculations and predictions. This development has significant implications for the AI/ML community, particularly in streamlining document processing workflows. Cloudflare envisions Clairic operating alongside LLMs and other models in a collaborative "System One" framework, where it can perform data extraction in conjunction with decision-making models like Clef and response-generating models like GLM-5.3. Notably, Clairic completes its multi-step operations in 19.3 seconds—less than a third of the time taken by GLM-5.3 alone while also cutting costs almost in half. While Clairic currently struggles with more complex calculation chains, it represents a promising step towards more efficient and cost-effective data processing solutions.