How much does Clairic cost?

Clairic costs $0.10 per million input tokens on Workers AI. Because the model generates no output tokens, there is no output charge — a difference from an LLM extraction call, where the written answer is billed too. Cloudflare's own claims pipeline example came out about 49% cheaper per million claims than the all-LLM version it replaced.

Is Clairic open source?

Clairic is not open-weight today. It runs only as a preview model on Workers AI under the id @cf/cloudflare/clairic. Cloudflare says it plans to release open weights later, after more improvements, which is the path it already took with the Clef decision models that are published under Apache-2.0.

How is Clairic different from Cloudflare's Clef models?

Clef answers typed questions — yes/no, choice, score — with a probability for each option. Clairic instead pulls values out of the input: text spans, numbers, dates, lists, estimates and forecasts. Cloudflare treats them as two model classes and shows them working together, with Clairic extracting the facts and Clef deciding what to do next.

What can Clairic not do yet?

Cloudflare lists real limits for Clairic: multi-step arithmetic and aggregation across many rows are weak, so it suggests doing that math in your own code. Estimates need supporting evidence in the input, the forecasting is basic and not a replacement for dedicated time-series models, and the 80% intervals can be too narrow on unfamiliar data.

How accurate are Clairic's forecasts?

In Cloudflare's 21-day test on TSA passenger counts, Clairic's forecasts were off by 3.3% on average, and 17 of the 21 actual counts fell inside its 80% interval. Cloudflare presents this as a basic forecasting ability for values a series implies, not as a substitute for a purpose-built forecasting model.

Cloudflare · 2026-10-19 · major

Clairic — Cloudflare's new model pulls typed fields out of documents

Clairic is Cloudflare's first extraction model: you name the fields you want, and it returns each one typed, with a confidence score, in one pass. It is in preview on Workers AI at $0.10 per million input tokens.

Cloudflare blog header for Clairic, its extraction model on Workers AI

Clairic reads a document once and hands back the exact fields you asked for, each with a confidence score.

Key specs

Price (input)	$0.10 / 1M tokens
Extraction step	0.46s vs 16.1s for GLM-5.3

Quick facts

Maker	Cloudflare
Model ID	@cf/cloudflare/clairic
Context window	65,536 tokens
Price	$0.10 / 1M input tokens, no output charge
Availability	Preview on Workers AI
License	Proprietary preview; open weights planned
Images per request	Up to 4 (PNG, JPEG, WebP)

Pricing

Input	$0.10 / 1M tokens
Output · Clairic generates no output tokens	$0

source ↗

What is it?

Clairic is an extraction model from Cloudflare, previewed on Workers AI on October 19, 2026. You pass it text or up to four images plus a list of fields — an order id, a total, a ship date — and it returns a typed answer for each one. Cloudflare calls it the first model in a new extraction class, separate from its Clef and Jev decision models.

How does it work?

Instead of writing an answer token by token, the model makes one parallel pass over the input and generates no output tokens at all, which is why only input is billed. Each field comes back with a type, a null probability and a confidence score; text answers carry character offsets into the source, and estimates return a median plus an 80% range inside bounds you set. Fields that do not fit one 8,192-token pass are split across parallel passes.

Why does it matter?

Pulling structured data out of invoices, emails and claims is usually an LLM call that is slow, charges for output and sometimes invents a value. In Cloudflare's damage-claim example, a Clairic-to-Clef-to-LLM chain finished in 19.3 seconds against 61.7 seconds for the all-LLM version, and cost about 49% less per million claims ($6,032 against $11,740). Confidence scores and source offsets also give developers something to route on when a field is uncertain.

Who is it for?

developers building document and agent pipelines

Frequently asked questions

How much does Clairic cost?: Clairic costs $0.10 per million input tokens on Workers AI. Because the model generates no output tokens, there is no output charge — a difference from an LLM extraction call, where the written answer is billed too. Cloudflare's own claims pipeline example came out about 49% cheaper per million claims than the all-LLM version it replaced.
Is Clairic open source?: Clairic is not open-weight today. It runs only as a preview model on Workers AI under the id @cf/cloudflare/clairic. Cloudflare says it plans to release open weights later, after more improvements, which is the path it already took with the Clef decision models that are published under Apache-2.0.
How is Clairic different from Cloudflare's Clef models?: Clef answers typed questions — yes/no, choice, score — with a probability for each option. Clairic instead pulls values out of the input: text spans, numbers, dates, lists, estimates and forecasts. Cloudflare treats them as two model classes and shows them working together, with Clairic extracting the facts and Clef deciding what to do next.
What can Clairic not do yet?: Cloudflare lists real limits for Clairic: multi-step arithmetic and aggregation across many rows are weak, so it suggests doing that math in your own code. Estimates need supporting evidence in the input, the forecasting is basic and not a replacement for dedicated time-series models, and the 80% intervals can be too narrow on unfamiliar data.
How accurate are Clairic's forecasts?: In Cloudflare's 21-day test on TSA passenger counts, Clairic's forecasts were off by 3.3% on average, and 17 of the 21 actual counts fell inside its 80% interval. Cloudflare presents this as a basic forecasting ability for values a series implies, not as a substitute for a purpose-built forecasting model.

Try it

env.AI.run("@cf/cloudflare/clairic", { model: "clairic", state, fields })