Source: blog.cloudflare.com/open-clairic
Cloudflare has released an updated Clairic, an extraction model it redesigned after the preview, along with Clairic-flash, a 9-billion-parameter version built for speed and a smaller footprint. The updated model reads PDFs and images, including tables, lists and dates, and accepts up to four images or PDF pages in one request. Both models are hosted on Workers AI, and their full open weights are published on Hugging Face.
The redesign changes how fields are generated: text and number fields are produced in parallel, with each field stopping at its own end marker, and special heads handle estimates, durations, calculated dates and explicit arithmetic. Responses still include confidence scores, character spans and probabilities, and the API schemas are unchanged from the preview endpoint. Cloudflare says responses should arrive in under one second, and it presents its accuracy figures as self-reported and directional.
Pricing has been reorganized, with the 9B model moved to the clairic-flash endpoint and the 27B model now served at clairic. Clairic is priced at $0.24 per million input tokens and Clairic-flash at $0.10, and output tokens are not billed. Until December 1, 2026, both models cost $0.10 per million input tokens under a promotional discount.
Why it matters if you build agents
Applications already using the preview Clairic endpoint receive the new extraction model automatically, since the update is rolled out in place. Teams can also run the open weights themselves or move cost-sensitive extraction work to Clairic-flash.
Developers should plan for the price change on December 1, 2026, when Clairic returns to $0.24 per million input tokens. Output tokens are not charged, so cost estimates should count only input volume.
Get stories like this every weekday
One short email with the day's agent tooling news.