---
title: open-jev demo · five System One models
emoji: ⚖️
colorFrom: yellow
colorTo: indigo
sdk: static
pinned: false
license: apache-2.0
short_description: Typed decisions in the browser on WebGPU, five models
models:
  - SupersonicLabs/Julia-1
  - SupersonicLabs/Julia-1-ONNX
  - fastino/GLiNER2.5-Decide
  - onnx-community/GLiNER2.5-Decide-ONNX
  - onnx-community/kev-0.6b-ONNX
  - onnx-community/kev-4b-ONNX
  - onnx-community/open-jev-deberta-v3-large-ONNX
datasets:
  - LocalLLaMA/typed-decisions
---

# open-jev demo, with five System One models

A preview build of [Nico Martin's open-jev demo](https://huggingface.co/spaces/nico-martin/open-jev-demo) with two models added: Fastino's [GLiNER2.5-Decide](https://huggingface.co/fastino/GLiNER2.5-Decide) ([ONNX](https://huggingface.co/onnx-community/GLiNER2.5-Decide-ONNX)) and Supersonic Labs' [Julia 1](https://huggingface.co/SupersonicLabs/Julia-1) ([browser export](https://huggingface.co/SupersonicLabs/Julia-1-ONNX)), next to Kev 0.6B, Kev 4B and open-jev. Everything runs locally in your browser on WebGPU.

- **Pick a model by link:** `?model=julia-1`, `?model=gliner2-decide`, `?model=kev-4b`, `?model=kev-0.6b`, `?model=open-jev` (or just `?julia-1`).
- **Benchmark page:** all five models on the 2,000 questions of [typed-decisions](https://huggingface.co/datasets/LocalLLaMA/typed-decisions) (test split, Apache-2.0), plus a button that reruns it on your own GPU. Open it from the top bar or link to it with `?page=benchmark`, e.g. `?model=julia-1&page=benchmark`.

Built from these open PRs:
- GLiNER2.5-Decide: [nico-martin/open-jev#1](https://github.com/nico-martin/open-jev/pull/1), [nico-martin/open-jev-demo#1](https://github.com/nico-martin/open-jev-demo/pull/1)
- Julia 1: [nico-martin/open-jev#2](https://github.com/nico-martin/open-jev/pull/2), [nico-martin/open-jev-demo#2](https://github.com/nico-martin/open-jev-demo/pull/2)

The app is Nico's; see [open-jev](https://github.com/nico-martin/open-jev) and [open-jev-demo](https://github.com/nico-martin/open-jev-demo) for the source. The benchmark data is the test split of [LocalLLaMA/typed-decisions](https://huggingface.co/datasets/LocalLLaMA/typed-decisions) (Apache-2.0).