Notes
Where to call Jev (September 2026 channel map)
Developer map of TypeSafe Jev early access, Vercel AI Gateway evaluate paths, and OpenRouter Decisions — with dates and primary links.
Unofficial notes. Not affiliated with TypeSafe AI. Prefer primary changelogs over press when you wire production.
Direct TypeSafe API
TypeSafe announced System One and Jev early access on 14 Sep 2026:
- Launch post: Introducing System One Models and Jev
- Docs: docs.typesafe.ai
- Native endpoint from our fact card:
POST https://api.typesafe.ai/v1/systemonewithmodellikejev-latest(see API quickstart)
Use this path when you want TypeSafe’s own client shapes (noul / choice / score) without a gateway in the middle.
Vercel AI Gateway
Two dated Gateway entries matter:
- 16 Sep 2026 — TypeSafe AI’s Jev now available on AI Gateway
- Model id:
typesafe-ai/jev - AI SDK 7
experimental_evaluate(evaluate) API - Changelog states Jev is free on AI Gateway until 25 Sep 2026 (promo end date from that post; re-check the live Gateway card before budgeting)
- Model id:
- 21 Sep 2026 — AI Gateway now supports TypeSafe clients and an HTTP API for Jev
- Point an existing TypeSafe client at
https://ai-gateway.vercel.sh/typesafe - Or
POST https://ai-gateway.vercel.sh/v1/evaluatewithmodel: "typesafe-ai/jev" - Billing and logs go through AI Gateway on those paths
- Point an existing TypeSafe client at
Gateway question naming on the evaluate HTTP/SDK path uses boolean / choice / score. TypeSafe-native clients keep noul when you use the TypeSafe-compatible base URL above.
OpenRouter Decisions API
OpenRouter listed Jev around 18 Sep 2026:
Public card facts we verified against the Jev 1.13 page: $0.042 / M input, $0 output, 32K context, Decisions API (not chat completions). Match that with jev-facts on this site — do not point a chat-completions SDK at System One.
Also covered (not primary sources)
Short press on the same launch week:
Pick a channel
| Need | Prefer |
|---|---|
| Official shapes + TypeSafe console keys | Direct api.typesafe.ai |
| AI SDK / Gateway budgets / existing Vercel stack | AI Gateway typesafe-ai/jev or /v1/evaluate |
| Multi-provider routing beside other models | OpenRouter Decisions (typesafe/jev-latest / typesafe/jev-1.13) |
For primitives and when to skip chat LLMs, see What is Jev AI and Jev AI vs LLM. Public examples are on Jev AI cases and demos. Clients and SDKs are in the GitHub project directory.