AI Agent News/stories/cloudflare-launches-two-system-one-models-clef-and-clairic-and-calls-itself-a.md

Infrastructure

Cloudflare launches two System One models, Clef and Clairic, and calls itself a neolab

Cloudflare has started launching its own models, Clef and Clairic, both System One models designed for fast decisions and data extraction. In an interview, the company's AI platform lead said Cloudflare may also build reasoning models and other LLMs, focused on post-training rather than pre-training from scratch.

Source: latent.space/p/cloudflare

Michelle Chen, who runs Cloudflare's AI Platform on the product side, called the company a neolab in a talk at the AI Engineer Code Summit. Cloudflare's first two models are Clef, a decision model modeled on TypeSafe AI's Jev, and Clairic, which Cloudflare describes as a new class of System One extraction model. Chen said Clef took about two weeks to build, while TypeSafe AI spent two years on Jev, and she credited TypeSafe for inventing the approach.

Clairic takes unstructured input and extracts, computes, estimates and forecasts typed values in one forward pass. Chen said the team targeted cases where LLMs were overkill or unreliable, such as pulling an order number from an invoice, and where LLM output could be malformed JSON. She also said Cloudflare's model work will center on post-training existing resources rather than building pre-trained models from scratch.

Why it matters if you build agents

Developers who send simple decisions or structured extraction tasks to a large LLM may find a smaller System One model faster and more reliable, since Chen says output is measured in milliseconds. Clairic is aimed at turning messy text into typed values without free-form generation. The article gives no pricing, availability or API details, so check Cloudflare's documentation before building on either model.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe