Cloudflare has announced a preview of Clairic, the first model in a new class of System One-style extraction models, hosted on Workers AI as @cf/cloudflare/clairic. The company says it intends to release the model open-weights soon. Clairic is designed to extract, compute, estimate, and forecast values from unstructured inputs at significantly higher accuracy, faster latency, and lower cost than traditional LLMs.
The announcement builds on the recent wave of interest around System One models, a concept drawn from Daniel Kahneman's book Thinking, Fast and Slow. Where System Two models (like LLMs) are built for deliberate reasoning, System One models are meant to be fast and precise. Cloudflare credits TypeSafe, which introduced Jev and prompted discussion of a new class of models that are not LLMs and how they fit into developer workflows.
- Clairic, the first extraction model, is available in preview on Workers AI as
@cf/cloudflare/clairic - Open-weights release is planned for the near future
- The model handles tasks including extraction, computation, estimation, and forecasting from unstructured inputs
- Cloudflare claims higher accuracy, faster latency, and cheaper cost compared with traditional LLMs
The pitch matters for developers because extraction and estimation tasks are common in production pipelines: pulling structured values out of messy text, estimating fields, or forecasting figures. Running those on a purpose-fit model, rather than prompting a general LLM, can reduce inference cost and improve response times without sacrificing precision. Because Clairic is hosted on Workers AI, teams can call it through the same platform already used for other inference workloads, and the planned open-weights release should give developers the option to self-host or fine-tune later. The preview is a chance to benchmark Clairic against current LLM-based extraction approaches before deciding where it fits.
Source: https://blog.cloudflare.com/clairic-extraction-model/