Small decisions. Interesting possibilities.Submit contentSubmit

deepseek-v4.1-flash-jev

you can make any open source model behave like jev with just a bit of inference engineering. it's shockingly easy. to prove it, we built a new endpoint we're calling deepseek-v4.1-flash-jev. see the demo below. here's how it's done: sglang (an inference engine) offers a scoring endpoint in addition to the normal…

Added to Jevfast

How it uses Jev

This is a Jev-inspired inference experiment, not a documented call to TypeSafe’s Jev API. The post describes an endpoint that scores candidate answers for an input and returns probabilities; its example ranks possible letter answers.

App workflow

  1. Send an input and candidate answers to the scoring endpoint

    Send an input and candidate answers to the scoring endpoint.

  2. Add a closing think tag for the described DeepSeek setup

    Add a closing think tag for the described DeepSeek setup.

  3. Receive probabilities for each candidate

    Receive probabilities for each candidate.

  4. Use the scores to compare candidate answers

    Use the scores to compare candidate answers.

Why it is interesting

It demonstrates that a scoring endpoint can expose choices and probabilities instead of streamed text generation.