Cloudflare introduces Clairic, a new extraction model that pulls structured data from unstructured inputs faster and cheaper than traditional large language models.
Key points
- Clairic is available in preview on Workers AI as @cf/cloudflare/clairic.
- The model extracts, computes, estimates, and forecasts values from unstructured text.
- It operates as a System One model, providing typed outputs without token generation.
Cloudflare has announced a preview of Clairic, a new class of System One extraction models hosted on Workers AI. This tool allows users to extract structured data from unstructured inputs with higher accuracy, lower latency, and reduced cost compared to traditional large language models.
What happened
Cloudflare released Clairic as the first model in a new category called System One extraction models. According to the company’s blog post, Clairic can extract, compute, estimate, and forecast values from unstructured inputs. The model is currently available in preview on Workers AI, and Cloudflare intends to release its weights as open-source soon.
Unlike previous System One models that acted as decision models, Clairic focuses on extracting direct values. It handles tasks such as finding invoice totals, delivery dates, or repair costs. These tasks often require structured outputs that traditional large language models handle inefficiently by generating tokens. Clairic provides answers in a single pass, typically responding in under a second.
Why it matters
System One models are designed for speed and precision rather than complex reasoning. They contrast with System Two models, which perform deliberate reasoning like standard large language models. System One tools are purpose-fit for specific, narrow tasks. They provide typed outputs and probabilities quickly. This approach reduces computational costs and latency. Teams use these models for repetitive data extraction. They avoid the variability and hallucination risks associated with generative text models.
The fix
- Review current workflows that use large language models for simple data extraction.
- Test Clairic on Workers AI for tasks requiring structured numerical or textual outputs.
- Monitor confidence scores provided by the model to validate extracted data accuracy.
- Plan integration of open-weights versions once Cloudflare releases them publicly.
Why it matters
Organizations often overuse large language models for simple data tasks. This approach increases costs and introduces unnecessary latency. Teams should evaluate whether a task requires reasoning or just extraction. Using specialized models for narrow tasks improves efficiency. It also reduces the risk of inconsistent outputs. Users must verify that the model’s typed outputs match their system requirements. Confidence scores help determine if human review is necessary. This shift allows for more scalable data processing pipelines.
What to do and how to stay safe: Cloudflare
- Audit existing applications to identify repetitive data extraction tasks handled by large language models.
- Evaluate whether these tasks require reasoning or can be solved with direct value extraction.
- Test the preview version of Clairic on non-critical workflows to assess accuracy and speed.
- Monitor vendor updates for the open-weights release and plan for local deployment if needed.
Step-by-step guide: What is password policies and how they protect systems
General security guidance from the Crypt Weekly newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is Clairic?
Clairic is a System One extraction model from Cloudflare that extracts, computes, estimates, and forecasts values from unstructured inputs.
How does Clairic differ from large language models?
Clairic answers in a single pass without generating tokens or reasoning, resulting in faster latency and lower cost.
Will Clairic be open source?
Cloudflare intends to release the model’s weights as open-source soon, according to their announcement.
