Jev on Cloudflare AI Gateway

It does sit in the path, since that is what a gateway does. This post is an availability announcement only and says nothing about logging, retention, or what is stored about prompts, so check the AI Gateway documentation before pointing sensitive state at it.

It is presented as Jev itself, the first System One model on the gateway, returning typed Noul, Choice, and Score answers. The post gives no model version, per-model pricing, or rate limits, and Cloudflare sets those, so verify current docs before depending on them.

Practically: send state plus typed questions and read typed answers back instead of parsing generated text. No code sample or function signature appears in the post, so the exact Workers binding shape has to come from the Cloudflare AI docs it links.

The post does not address authentication at all. Gateways typically authenticate with their own credential plus a provider key, but this announcement mentions neither, so treat the auth path as undocumented and check the model page in the Cloudflare AI docs first.

The announcement is framed around Workers: it invites you to try the first System One model on AI Gateway and points at Cloudflare's AI docs for Jev. It does not describe calling the gateway from other runtimes or show a binding, so Workers is the documented path and anything else is unstated.

There is a linked docs page. The post points to a Cloudflare AI documentation page titled Jev (typesafe), which is where the model's request and response details live. The tweet itself is an availability announcement and gives no parameters, pricing, or rate limits of its own.

State plus typed questions, and you get structured answers back that your code can use directly. That is the shape Cloudflare describes rather than free-form prompting: you supply the situation and the questions you want answered, and the model returns typed answers instead of generated prose.