Cloudflare previews Clairic, a model that pulls typed fields such as invoice totals and dates from text in one pass and costs no output tokens

Cloudflare announced Clairic on 19 October, a preview model on Workers AI for reading values out of unstructured text and up to four images. You list fields, such as an invoice total or delivery date, and each comes back typed, with a confidence score and the exact position in the source text. Missing values return null. It can also estimate and forecast numbers with an 80% range. Cloudflare says it is faster and cheaper than language models and will release open weights soon.

• List the paperwork your team retypes every week, such as invoices, orders or complaint emails, and write down the five fields you copy out of each.
• If someone already automates this with a chatbot, ask them to log how often it gets a field wrong, since a confidence score is only useful when you compare it with that.

Source: Cloudflare Blog. This is our short summary of their reporting. The paragraphs marked "Our read" and the things to do are SeedFoundry's opinion. The summary is drafted with AI tools. Image: Cloudflare Blog, from its own announcement. How we source and credit