Jev Playground · jevplayground.com

It says the real thing. The page states it runs the actual TypeSafe Jev model, that answers are not simulated, and that no signup is needed. It also says free tries have usage limits, so expect throttling before you have evidence either way.

The site never says. It describes itself as an independent third-party tool, not affiliated with or endorsed by TypeSafe AI, and links TypeSafe's official site and Vercel's model listing as the official sources. Your state text goes to whatever backend the site operates, so treat it as a sandbox.

The state field shows a counter against an 8,000-character cap, so long documents get truncated by you rather than by the model. The Choice example starts at 4 of 10 options with an add-choice control, meaning the option ceiling is enforced in the interface itself.

That is the sensible use. It ships editable examples for support routing, refund verification, incident urgency, and agent next action, plus a browser agent, citation verifier, email triage, model router, code review, and a tool-call risk gate. Load your own wording and see if the decisions hold.

Loaded examples are Support Routing (Choice), Refund Verification (Boolean), Incident Urgency (Score) and Agent Next Action (Choice), with further tasks including browser agent, citation verifier, email triage, model router, code review and a tool-call risk gate.

Yes, it offers single-question and multiple-question modes, and the loaded cases use choice, boolean and score decisions so you can see each primitive in the same interface.

The site has its own Pricing page and says free tries have usage limits, but it is an independent tool rather than the TypeSafe console, so treat its rate and limit statements as the operator's own.