Service availability

Current development preview

UseJevAI is an independent interface for exploring Jev AI. It is not the official TypeSafe AI website. This page is a preview of service information; final commercial terms have not been issued.

Available features

You can browse templates, edit questions, copy configurations, prepare a batch file, and compare proposed token packs. A configured local environment can make live Playground calls through OpenRouter.

Playground, Batch, and API runs depend on current service availability, shown on each page. Your account tracks input-token usage and saved results.

PayPal checkout is available only for administrator Sandbox tests with virtual funds. Real payments are not open.

Model results

Model outputs can be incorrect or incomplete. Probabilities describe model estimates and do not guarantee the correctness of an individual answer. Review results against your own examples and application requirements.

Proposed purchases

The pricing page shows one-time input-token packs. When Sandbox checkout is unavailable, the dialog only previews the pack. A locally enabled administrator test can proceed to PayPal Sandbox using virtual funds. For refund requests, open a support ticket with your order ID and reason. Requests are reviewed individually by an administrator; submitting a ticket does not hold tokens or automatically issue a refund. Approved unused-token refunds use the original order rate. Your statutory rights are unaffected. Live payment remains closed.

Token validity and usage limits

The 20K trial allowance expires seven days after activation. Purchased tokens never expire. Free runs share a site-wide model-cost budget of $2 per day, resetting at 00:00 UTC. An administrator may increase that day's budget; the default applies again the next day. Trials can also pause when the cumulative budget is reached.

Paid usage allows up to 10 concurrent runs and 60 runs per minute per account, shared across Playground, API and Batch. Each run accepts up to 32K total billable input tokens and 32 questions; each paid Batch accepts up to 100 rows. Each row counts as a run.