Cloudflare launches Clef and Clef-flash, decision models with typed outputs and probabilities
In one sentence On October 1 Cloudflare introduced Clef and Clef-flash, models that do not generate text but return typed decisions with probabilities: built for classification, routing and fast decisions in agent workflows, they run on Workers AI and ship Apache 2.0 weights on Hugging Face.
Many applications do not need an AI model to write text: they just need it to decide. Is this request urgent? Which team should handle it? Is this domain suspicious? Using a large generative model to answer yes or no is slow and expensive, and the answer then has to be parsed.
Cloudflare introduced Clef and Clef-flash, two models built specifically for this: they take text (or, for Clef, also an image) plus a question and return an answer in a fixed format, such as a choice among options or a score, together with its probability. Clef is the more precise version, Clef-flash the one built for speed.
The models run on Cloudflare's Workers AI platform, but the weights are published on Hugging Face under Apache 2.0, so they can also be self-hosted. Cloudflare says it does not read, store or train on customer requests unless the customer opts into fine-tuning.
Companies
Cloudflare
Tools
Clef, Clef-flash, Workers AI
Tags
Sources