Orgs cut costs with Cloudflare's Clairic, a System One model for fast unstructured data forecasting

Cloudflare has announced a preview of Clairic, a new extraction model designed to process unstructured data more efficiently than traditional large language models. The company states that this System One class model can extract, compute, estimate, and forecast values from inputs with significantly higher accuracy, faster latency, and lower cost. It is currently hosted on Workers AI, with plans to release the model weights openly in the near future.

Clairic represents a shift away from the reasoning-heavy approach of standard large language models. According to Cloudflare, it answers queries in a single pass without generating tokens, which means it cannot produce a mistyped answer or “hallucinate” in the same way as an LLM. The model returns typed values such as names, dates, and lists, along with confidence scores. This allows organisations to trust the output or flag it for manual review, ensuring reliability in structured data tasks.

What we know

The concept of System One models was popularised by TypeSafe a month ago, drawing on Daniel Kahneman’s distinction between fast, precise thinking and deliberate reasoning. While previous System One models like Jev and Clef focused on decision-making and probabilities, Clairic is the first in a new class dedicated to extraction. It handles tasks such as identifying invoice totals, delivery dates, or traffic forecasts, which do not fit neatly into predefined probability lists.

Cloudflare notes that many structured-output calls currently burden frontier large language models with repetitive tasks. Clairic addresses this by returning numbers within an eighty per cent range, or pulling typed values directly from the input. It also provides estimated values based on evidence in the input and forecasted values for future series steps. Unlike large language models that take several seconds to reason through a step, Clairic typically responds in under a second.

Defender's checklist

• Verify that data extracted by Clairic meets your organisation’s accuracy requirements before integration.
• Review the confidence scores provided with each output to determine if manual verification is needed.
• Ensure that the model’s input data is clean and consistent to maximise extraction accuracy.
• Monitor the model’s performance in preview to identify any edge cases or failures in value extraction.

The background in plain language

Teams usually rely on large language models for complex reasoning, but these models can be slow and prone to errors when handling simple data extraction tasks. System One models offer a faster, more precise alternative for specific use cases. They process inputs in a single pass, returning typed values without the overhead of token generation. This approach reduces costs and latency, making it suitable for high-volume data processing tasks where speed and accuracy are critical.

Background: Large language models

Large language models are zero-shot reasoners only in appearance. They predict the next word by matching patterns in training data, not by applying logical rules. This means they can solve novel problems if similar examples exist in their data, but they fail when true novel logic is required.

Read the full guide: Are Large Language Models Zero-Shot Reasoners? The Reality Check

What to do and how to stay safe: Cloudflare

• Audit your current data extraction workflows to identify tasks that could benefit from System One models.
• Test Clairic with sample data to evaluate its accuracy and latency compared to your existing tools.
• Establish clear criteria for when to trust the model’s output versus when to trigger manual review.
• Stay updated on the release of open-weights versions to explore self-hosting options for greater control.

General security guidance from the TechWire Beat newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What is Clairic and how does it differ from large language models?

Clairic is a System One extraction model that extracts, computes, estimates, and forecasts values from unstructured inputs without generating tokens or reasoning, offering faster and cheaper processing than large language models.

When will Clairic be available as an open-weights model?

Cloudflare intends to release the model open-weights soon, but has not specified an exact date for the release.

What types of outputs does Clairic provide?

Clairic provides extracted text spans, computed numbers and dates, estimated values with confidence intervals, and forecasted future values, all accompanied by confidence scores.