Introducing Clairic, a new System One class of extraction models

Cloudflare has introduced Clairic, a preview System One extraction model on Workers AI designed to extract, compute, estimate, and forecast typed values from text and up to four images in a single parallel pass, with open weights planned for a later release. Unlike generative LLMs, Clairic returns structured outputs such as text spans, numbers, dates, lists, probability distributions, and forecasts, along with confidence scores, null probabilities, source offsets where applicable, and 80% uncertainty intervals for estimates and forecasts. Cloudflare positions it alongside decision models such as Clef and generative LLMs: Clairic converts unstructured inputs into facts, decision models select actions from those facts, and LLMs generate natural-language responses. Examples include extracting invoice and support-email details, calculating totals and elapsed time, reading receipt images, estimating home prices, and forecasting TSA checkpoint traffic; Cloudflare reports a 3.3% average error in a 21-day TSA forecasting test and claims a Clairic–Clef–LLM workflow was faster and less expensive than an all-LLM pipeline. The company notes that the early model is strongest at reading and short calculations, while complex arithmetic, broad aggregation, unsupported estimates, and detailed long-term forecasting remain limitations, recommending users rely on confidence signals and human review where appropriate.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.