Cloudflare Previews Clairic, 49% Cheaper on Claims Pipeline
TL;DR
- Clairic generates zero output tokens and bills input-only at $0.10 per million, structurally undercutting GPT-class extraction pricing on cost per task.
- A three-stage Cloudflare pipeline (Clairic + Clef + LLM) cut a test insurance-claims workflow from 61.7 to 19.3 seconds, a 69% latency reduction.
- Cloudflare's 'System One' label is a taxonomy play: positioning the fast, non-reasoning extraction tier as a named category before any competitor defines it.
19.3 seconds versus 61.7 seconds. That is Cloudflare's headline comparison in a preview announcement for Clairic, a new extraction model it is launching on Workers AI as `@cf/cloudflare/clairic`.
"Clairic is an extraction model that is able to extract, compute, estimate, and forecast values from unstructured inputs," the Cloudflare post by Michelle Chen reads. The framing is that extraction belongs to a different class of model than conversational LLMs. "A System One model should be fast and precise, which contrasts a System Two model that is capable of deliberate reasoning."
The showcase is a damaged-product insurance claim pipeline. Clairic pulls typed values out of the customer email, hands them to Clef, a decision model, for approval, and then an LLM writes the customer response. Against an all-LLM baseline, Cloudflare says "the cascade finished in 19.3 seconds against 61.7 seconds: 69% less time, or about 3.2 times faster." Cost per million claims comes to $6,032 against $11,740.
On forecasting, Cloudflare ran Clairic on 21 days of TSA passenger counts. "Across the 21 days, Clairic's forecasts were off by 3.3% on average, and 17 of the 21 actual counts landed inside its 80% interval."
Cloudflare is upfront that the model is early. "Reading is strong, but multi-step arithmetic isn't yet," the post says. Forecasting is "basic." The baseline LLM in the comparison is not named. "We intend to release the model with open weights soon," Chen writes.
What others are reporting
-
Cloudflare Developers (changelog) Read →
First-party technical changelog with full type table, $0.10/M input pricing, API code samples, and explicit preview-stage limitation disclosures absent from the announcement blog post.
Clairic handles these. You send a state and a schema of typed fields, and it returns one typed value per field.
-
Unite.AI Read →
Places Clairic alongside TypeSelf's Jev and Cloudflare's Clef in a three-model pipeline, framing extraction as a named category distinct from decision and generative models.
The model answers every field in a single parallel forward pass, with no reasoning step and no generated tokens.
-
LavX News Read →
Frames the release as a shift from monolithic LLMs toward specialized pipelines, and raises skepticism on whether 'System One' is durable taxonomy or marketing positioning.
Extraction models turn raw input into typed values, decision models turn values into decisions, and LLMs turn decisions into words.
-
ExplainX.AI Read →
Audits Cloudflare's benchmark numbers and surfaces a 7.3-second discrepancy between published step times and the claimed all-LLM baseline total, separating verified data from assertion.
The cascade column adds up, but the all-LLM column does not.
-
GIGAZINE Read →
Mainstream international tech outlet providing concrete end-to-end pipeline comparison numbers and accessible framing of the extraction-versus-generation distinction for a broad audience.
By eliminating the step of generating the answer as a sentence, it can reduce processing time and costs.
-
CellCog Read →
Foregrounds Cloudflare's own candid limitation disclosures, especially on multi-step arithmetic, as an operational constraint teams must weigh before adopting Clairic in production.
Reading is strong, but multi-step arithmetic isn't yet.
Shared on Bluesky by 3 AI experts
-
Tim Kellogg @timkellogg.me
Clairic: a system one model for regression like Jev, except it generates continuous numbers e.g. start & end indices to slice/quote a string value out of a document blog.cloudflare.com/clairic-extr...
View on Bluesky → -
Introducing Clairic, from Cloudflare. A System One extraction modeI, capable of extracting, computing, estimating, and forecasting data from unstructured inputs. https://cfl.re/3UcTaMP
View on Bluesky →
Originally reported by blog.cloudflare.com
Read the original article →Original headline: Cloudflare Launches Clairic, a 'System One' Extraction Model 3.2x Faster and 49% Cheaper Than LLMs