Clairic brings typed extraction to Workers AI
Clairic separates structured extraction and estimation from free-form generation, returning typed answers with source spans or uncertainty information for applications that need machine-consumable values.
Logged at IST: 2026-10-20 09:19 IST
What it is: Cloudflare's preview announcement for Clairic, a hosted Workers AI model for extracting, computing, estimating, and forecasting typed values from unstructured inputs.
Gist: A request supplies text (and optionally images) plus a schema of named fields. Clairic can return text spans with offsets, typed values such as numbers and dates, estimates with a median and 80% interval, and forecasts. Cloudflare says it runs in a single forward pass without generating output tokens; its docs list input-token pricing of $0.10 per million tokens. The model is in preview on Workers AI, and Cloudflare says it intends to release open weights soon.
Why it matters: For extraction or bounded prediction, a typed answer with provenance or uncertainty may be easier to validate and cheaper to integrate than asking a general LLM to emit free-form JSON. Cloudflare's speed, accuracy, and cost comparisons are vendor claims; the model can still make mistakes, and the announcement is a preview rather than independent evaluation.
Source: https://blog.cloudflare.com/clairic-extraction-model/ Docs: https://developers.cloudflare.com/workers-ai/models/clairic/ Saved link: https://x.com/Cloudflare/status/2112244042658423149