Full guide: What is a System One extraction model? Extraction models explained
You send the model a state and a set of named fields, each with a type such as number or date. It answers every field in one parallel pass and writes no tokens. A value is read from the input, computed from it, or predicted beyond it.
The difference from a decision model is the output space. A decision model picks from options you list. An extraction model fills in values you could not list in advance, such as an invoice total or a delivery date. Cloudflare’s summary: “Extraction models turn raw input into typed values, decision models turn values into decisions, and LLMs turn decisions into words.”
A field that asks for a single value returns a value of the declared type or null, and p_null gives the probability that the value is missing. A list with no matches comes back empty, and estimates and forecasts always return a prediction. Text answers come back as a span of the input.
Clairic is the only one so far, in preview on Workers AI. TypeSafe AI, which coined System One, had not commented on the new class as of 19 October 2026. The extraction model guide covers it in full.