Cloudflare previews Clairic extraction model that skips token generation
Original title: Cloudflare bets you’re overpaying an LLM to read your invoices
In one claims workflow, a Clairic pipeline finished in 19.3 seconds and cost 49% less than an all-LLM version.
Cloudflare released a preview of Clairic, an extraction model on Workers AI that returns typed fields from text and up to four images without generating tokens. It plans to publish the weights and charges only for input. In a damaged-product claims test, a Clairic cascade took 19.3 seconds and cost $6,032 per million claims, versus 61.7 seconds and $11,740 for a version that used an LLM at every step. Cloudflare calls the endpoint an early checkpoint and says longer arithmetic chains are unreliable.
Unconfirmed: Speed, cost and accuracy figures are from Cloudflare’s own tests, not an independent benchmark.
- The model id is @cf/cloudflare/clairic; weights are planned for publication.
- Clairic extracted claim facts in 0.46 seconds; Clef chose refund or repair in 0.22 seconds.
- GLM-5.3 needed 38 seconds for those two steps in Cloudflare’s comparison.
- Airport-count forecasts missed by 3.3% on average over 21 days.
- Cloudflare says code, not Clairic, should handle longer dependent arithmetic.
- Vendors
- Cloudflare
- Products
- Workers AI
From the original article
3 min read at helpnetsecurity.com
Cloudflare released a preview of Clairic, an extraction model that pulls typed values out of messy text and images in one pass without generating tokens. In Cloudflare’s test on a damaged-product support claim, a pipeline built around Clairic finished in 19. 3 seconds, against 61.
By Anamarija Pogorelec · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.