Cloudflare announced Clairic, a System One extraction model. It extracts, computes, estimates, and forecasts values from unstructured inputs. Clairic runs on Workers AI as @cf/cloudflare/clairic in preview. It returns typed values, spans, and confidence scores in one pass. It charges only for input tokens. It handles text and up to four images per request. Outputs include extracted text, computed numbers, estimated distributions, and forecasted series. It cannot hallucinate like LLMs but may still make mistakes. Cloudflare plans to release Clairic as open-weights soon. It is faster, cheaper, and more accurate than LLMs for extraction tasks.
What changed
LLMs handled extraction via token generation with higher latency, cost, and error rates.
What it unlocks
Extracting, computing, estimating, and forecasting typed values from unstructured text or images in one pass.
- 80% confidence interval
- single forward pass
- no output token cost
What you need to act on it
- preview access on Workers AI
- model identifier @cf/cloudflare/clairic
Sources