Cloudflare has released a preview of Clairic, the first model in what it calls a new class of “System One extraction models.” Instead of generating text, Clairic pulls typed values — numbers, dates, durations, lists, estimates, and forecasts — out of unstructured text and images in a single forward pass, with no output tokens at all.
The model is available now on Workers AI as @cf/cloudflare/clairic, callable from a Cloudflare Worker or through the REST API. A developer submits input text (plus up to four images) and a list of named fields, each with a type. Answers come back as values of that type, or as null when the input simply does not contain the answer. Text outputs include exact character offsets pointing back to the source; scalar outputs carry a confidence score and p_null, the model’s estimate that the value is missing.
In Cloudflare’s own test, a claims pipeline built around Clairic, its Clef decision model, and GLM-5.3 cost 49 percent less and ran about 3.2 times faster than GLM-5.3 alone. The official docs list a 65,536-token context window and pricing of $0.10 per million input tokens — and since the model generates no output tokens, billing is input-only. Cloudflare plans to release the weights openly after more checkpoint work, though no date was given.
Clairic targets the invoice-reading, form-parsing, and support-ticket jobs developers currently hand to large language models like ChatGPT, Gemini, and Perplexity — calls that can emit broken JSON or misread a digit while writing it out. Finance teams dealing with exactly this workload should see our best AI tools for accountants in 2026.
Source: Cloudflare Blog; Help Net Security
