Cloudflare Introduces Clairic, a New System One Extraction Model
Cloudflare has announced a preview of Clairic, its first System One-style extraction model designed to pull, compute, estimate, and forecast structured values from unstructured inputs.
Before you start: versions, package names and storage identifiers change over time. Check the commands against the official documentation for your setup before running them, especially anything that creates, deletes or overwrites data.
A New Class of System One Models
Cloudflare has announced a preview of Clairic, its first System One-style extraction model hosted on Workers AI under the identifier @cf/cloudflare/clairic. Available immediately for developers on Cloudflare Workers, the model is designed to extract, compute, estimate, and forecast values from unstructured inputs with higher speed and lower cost than traditional large language models.
Following the recent introduction of System One decision models like Jev and Clef, Clairic expands the model architecture to handle regression and extraction tasks. Unlike LLMs that process text token by token through slow, deliberate reasoning steps, Clairic operates in a single forward pass without generating output tokens.
Capabilities and Structured Outputs
Clairic processes unstructured text and up to four optional images alongside a map of typed fields. It returns strongly typed answers for each requested field in a single parallel pass. The model supports several output types:
- Extracted values: Text copied directly from the source, returned as character offset spans with confidence scores.
- Computed values: Numbers, dates, times, durations, and lists derived or calculated directly from the text following strict formatting rules.
- Estimated values: Predicted point values accompanied by a median, an 80% confidence interval, and specified percentile bounds.
- Forecasted values: Time-series predictions providing sequential future estimates with individual confidence intervals.
Every scalar response includes a probability of null value (p_null) and a confidence metric, ensuring that missing or unstated data points return a null value rather than a hallucinated guess.
Availability and Future Release
Clairic is currently available in preview on Cloudflare Workers AI. Cloudflare has also confirmed intentions to release the model weights publicly in the near future, allowing developers to self-host and integrate the extraction architecture into custom workflows alongside decision models and LLMs.
Source
Official announcement or documentation
Some links on this page may be affiliate links. If you buy through them we may earn a commission at no extra cost to you. See our affiliate disclosure.
