Skip to content
  • Models
  • Rankings
  • Ori
Sign Up
Sign Up

Jev Lab

Recipes that run live against TypeSafe's Jev through OpenRouter. Each one is the code plus the result.

Jev model page

  • Support message triage

    Five yes-or-no questions about every one of 95 messages, in about a second.

    Uses
    Jev
    Speed
    475 answers in 1.2 s
    Cost
    $0.0014 a run

  • Agent action approval

    Four checks on each of 24 tool calls in one request. Risky steps stop for a person.

    Uses
    Jev
    Speed
    96 answers in 0.5 s
    Cost
    $0.0004 a run

  • Extraction without guessing

    Jev picks each field from candidates found in the text, so it cannot invent a value.

    Uses
    Jev
    Speed
    12 fields in 0.6 s
    Cost
    $0.0002 a document

  • Prompt to questions

    Paste a prompt. Claude rewrites it as Jev questions, then Jev runs them on every row.

    Uses
    Jev + Claude Haiku 4.5
    Speed
    380 answers in 2 s
    Cost
    $0.0025 a run

  • Feed filter in plain words

    Write a rule in plain words and Jev flags every matching post, in one request.

    Uses
    Jev
    Speed
    40 posts in 0.5 s
    Cost
    $0.0003 a run

  • Voice assistant gated by Jev

    Coming soon

    A voice model hears the room. Jev decides, per sentence, if it was spoken to and what to do.

    Uses
    Jev + gpt-realtime-2.1
    Speed
    10 answers in 260 ms
    Cost
    $0.00005 a decision

  • Live chase opponent

    You run, Jev chases. Every chaser move is one request, two to three times a second.

    Uses
    Jev
    Speed
    2 to 3 moves a second
    Cost
    $0.16 an hour

  • Checkout form that notices

    A checkout form asks eight questions after every click and reacts to the answers.

    Uses
    Jev
    Speed
    8 answers in 300 ms
    Cost
    $0.0001 a check