Cloudflare Launches Clairic for Fast, Low-Cost Data Extraction
Cloudflare introduces Clairic, a System One model that extracts and forecasts data from unstructured inputs faster and cheaper than traditional LLMs.
Key points
- Clairic is the first System One extraction model, available in preview on Workers AI.
- It processes inputs in under a second without generating tokens or reasoning.
- The model returns typed values, computed numbers, estimates, and forecasts with confidence scores.
Cloudflare has announced a preview of Clairic, a new type of artificial intelligence model designed to extract and forecast data from unstructured inputs with greater speed and lower cost than traditional large language models.
The risk
Traditional large language models are often inefficient for tasks that require structured outputs, according to Cloudflare. The company notes that many current workflows send repetitive structured-output requests to frontier models for questions like invoice totals or delivery dates. These tasks do not have answers that can be listed in advance, which is needed for decision models to give probabilities. This inefficiency led Cloudflare to develop a new class of models specifically for extraction tasks.
Who is exposed
Clairic is available hosted on Workers AI as @cf/cloudflare/clairic, according to Cloudflare. The company intends to release the model open-weights soon. The model is designed for tasks where large language models are inefficient, such as extracting specific values or computing numbers from unstructured inputs. It targets users who need fast, typed, and cheap answers that are directly extracted, computed, or predicted from input data.
Cost and compliance
Clairic operates differently from decision models like Jev and Clef, which provide probabilities against listed questions. Instead, Clairic provides direct values extracted, computed, or predicted from input. Because it does not generate tokens, it charges only for input and typically responds in under a second. This contrasts with large language models that may take several seconds to reason through similar steps. The model outputs include confidence scores to help users evaluate the reliability of the results.
What should you do?
Review current workflows that use large language models for simple data extraction or numerical computation. Evaluate whether replacing these calls with a specialized extraction model could reduce latency and costs. Monitor the Cloudflare Workers AI documentation for the open-weights release to assess integration feasibility for your specific application requirements.
The background in plain language
System One models are designed for fast, precise tasks, contrasting with System Two models that perform deliberate reasoning. In this context, extraction models handle specific outputs like numbers or typed values rather than open-ended text. Teams usually deploy these specialized models to handle repetitive data tasks efficiently. Check whether your current AI usage involves simple data retrieval that could benefit from a faster, purpose-built tool. This approach reduces reliance on heavier models for tasks that do not require complex reasoning capabilities.
What to do and how to stay safe: Cloudflare
- Audit your application logs to identify API calls to large language models that request simple structured data like dates, names, or totals.
- Test the Clairic preview on Workers AI with a subset of your unstructured data to compare latency and output accuracy against your current solution.
- Implement confidence score thresholds in your code to automatically flag low-confidence extraction results for manual review or fallback processing.
- Plan for the open-weights release by documenting your current model integration patterns to facilitate a smoother transition once the weights are public.
Step-by-step guide: Browser security settings are the architectural walls of your digital life
General security guidance from the Vector Tribune newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the difference between Clairic and traditional LLMs?
Clairic is a System One extraction model that answers with direct values in a single pass without reasoning or generating tokens, whereas LLMs generate tokens and reason through steps, taking longer and costing more.
What types of outputs does Clairic provide?
Clairic provides extracted text values, computed numbers and dates, estimated values based on evidence, and forecasted future values, all accompanied by confidence scores.
Is Clairic available for immediate use?
Yes, Clairic is available in preview hosted on Workers AI, and Cloudflare intends to release the model open-weights soon.
