OneJev-4B

OneJev-4B is the 4B model of OneJev, a multimodal System One decision model. Give it a screenshot, a photo, a video or plain text along with a few typed questions, and it returns a calibrated probability for every option in a single forward pass. It is a full fine-tune of Qwen/Qwen3.5-4B on 99,193 questions drawn from real agent runs, videos and images.

Results

Accuracy in percent. The OneJev test set holds questions of the training kinds that no model saw in training. Jev 1.13's scores are its published ones, and it reads text only.

On one H200 with a 1280x720 screenshot, OneJev-4B answers 1 question in 64 ms and 10 questions in one request in 104 ms (10.4 ms per question).

Quick start

pip install "qev[torch] @ git+https://github.com/OmniJev/OneJev.git"
qev serve --model OmniJev/OneJev-4B

from qev import Client, Choice, Noul
from qev.media import data_uri

r = Client("http://localhost:8000").system_one(
    state={"task": "Pay the open invoice from ACME", "screen": "<image:1>"},
    media=[{"type": "image", "data": data_uri("screenshot.png")}],
    questions={"done": Noul("The invoice has been paid"),
               "next": Choice("What should the agent do next?", {"click": "click an element", "stop": "stop"})},
)

The server speaks TypeSafe's System One API plus a media field for images and video. More examples, the latency benchmark and the code are on GitHub.

OneJev-4B also runs on llama.cpp from the GGUF build by mradermacher, without PyTorch. Images work there; video needs the PyTorch server.

brew install llama.cpp
pip install git+https://github.com/OmniJev/OneJev.git
qev serve --gguf mradermacher/OneJev-4B-GGUF:Q8_0

All sizes

All sizes are in the OneJev collection.

Citation

@misc{onejev2026,
  title        = {{OneJev}: A Multimodal System One Decision Model},
  author       = {{OmniJev Team}},
  year         = {2026},
  howpublished = {\url{https://github.com/OmniJev/OneJev}}
}

License

Apache 2.0