Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

openjev

Open JEV: local bilingual (English/Chinese) choice and binary decisions read from a frozen Qwen3-4B's next-token label distribution.

Projectresearch
View on GitHub ↗

Overview

Open JEV is an independent research preview that gives local, bilingual probability decisions. You supply English or Chinese context, free-form questions and your own candidate descriptions, and it returns a choice and relative probabilities only over those candidates, with no inserted abstention option. It supports Choice and Binary decisions, multiple questions per request, editable examples and JSON export through a web interface that switches between English and Chinese.

The current backend is a frozen Qwen3-4B-Instruct-2507 read through one next-token answer-label distribution per question, not a newly trained model, and the README states there is no demonstrated performance lead over Jev or others. Questions run sequentially and re-encode the context; shared-context computation is planned. It publishes model identity, scoring protocol, limitations and a frozen results snapshot, and returns stable candidate IDs that an application can map to workflow branches. It is not affiliated with TypeSafe.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich supplied candidate answers this question for the context?—
Yes / noBinary question about the contextyes · no

How it works

  1. Enter context and questions with candidate descriptions.
  2. Read the next-token label distribution for each question.
  3. Return candidate IDs with probabilities.
  4. Map the selected ID to an application branch.

Strengths

  • Runs locally with no paid API.
  • Bilingual interface and inputs.
  • Transparent about limitations and scoring protocol.

Caveats

  • Research preview; no performance lead claimed.
  • Sequential questions re-encode context.
  • No abstention option.

At a glance

Type
Project
Status
research
Language
Python
Runtime
local web app
Has code
yes

Use cases and integrations

How curated lists describe it

  • Open research: independent local preview that answers bilingual probability questions from context, questions, and candidate answers, inspired by TypeSafe Jev.

    yibie/awesome-jev

  • Local bilingual probability decisions from context, questions, and candidate answers.

    cobanov/awesome-jev

  • Local bilingual probability decisions from context, questions, and candidate answers. Independent research preview inspired by TypeSafe Jev.

    hellogumbo/awesome-jev

Frequently asked questions

Which model does it use?

A frozen Qwen3-4B-Instruct-2507.

Does it beat Jev?

The README says no performance lead is demonstrated.

Which languages are supported?

English and Chinese.

Related projects

  • brier

    Mangaba AI's local, calibrated typed-decision model for Portuguese text, trained to optimize the Brier score; proprietary since v5.

    ProjectproductionPython
  • blink★ 22

    Open one-pass typed-decision model with a dependency-free C99 runtime that also builds to a 66 KB WebAssembly module.

    ServicebetaC
  • Foq★ 11

    Local open alternative to Jev using a 1.58-bit ternary 8B model to answer booleans, choices, scores and Pydantic extraction in ~25 ms.

    ProjectbetaPython
  • browser-laya

    Runs the open Laya decision model entirely in a browser tab with WebGPU and ONNX Runtime Web: no server, no API key.

    ServicebetaTypeScript
  • ifReflex.cpp

    llama.cpp-based C++ server that turns any LLM into a System One decision model, reading answers from next-token logits.

    ProjectbetaC++
  • jevalt

    Three open 4B decision models for English, Turkish and German that serve the Jev /v1/systemone API on a CPU in about 3 GB of RAM.

    ServicebetaPython