---
title: OneJev1.6-0.8B-GGUF
canonical_url: "https://www.modelscope.ai/models/OmniJev/OneJev1.6-0.8B-GGUF"
md_url: "https://www.modelscope.ai/models/OmniJev/OneJev1.6-0.8B-GGUF.md"
repository: OmniJev/OneJev1.6-0.8B-GGUF
last_updated: 2026-10-18
license: apache-2.0
pipeline_tag: image-text-to-text
tasks:
- image-text-to-text
base_model:
- OmniJev/OneJev1.6-0.8B
base_model_relation: quantized
library_name:
- gguf
downloads: 44
stars: 0
tags:
- onejev
- system-one
- decision-model
- calibration
- multimodal
- gui-agent
- gguf
- llama.cpp
---
# OneJev1.6-0.8B-GGUF
> OneJev1.6-0.8B-GGUF - An open-source model by OmniJev on ModelScope. This repository holds GGUF files of OneJev1.6-0.8B for llama.cpp.
OmniJev/OneJev1.6-0.8B-GGUF is a image-text-to-text model on ModelScope. licensed under apache-2.0. derived from OmniJev/OneJev1.6-0.8B.
- **Repository**: OmniJev/OneJev1.6-0.8B-GGUF
- **License**: apache-2.0
- **Tasks**: image-text-to-text
- **Base model**: OmniJev/OneJev1.6-0.8B
- **Tags**: onejev, system-one, decision-model, calibration, multimodal, gui-agent, gguf, llama.cpp
- **Downloads**: 44
- **Stars**: 0
- **Last updated**: 2026-10-18
Source: https://www.modelscope.ai/models/OmniJev/OneJev1.6-0.8B-GGUF
---
This repository holds GGUF files of [OneJev1.6-0.8B](https://modelscope.cn/models/OmniJev/OneJev1.6-0.8B) for llama.cpp.
OneJev1.6-0.8B is the 0.8B model of OneJev 1.6, 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 averages the weights of three full fine-tunes of [Qwen/Qwen3.5-0.8B](https://huggingface.co/Qwen/Qwen3.5-0.8B), trained on 390,000 to 440,000 questions drawn from real agent runs, videos, images and the training splits of public decision datasets.
## Files
Converted with [llama.cpp](https://github.com/ggml-org/llama.cpp/commit/23f7aabc16c72f6f707dba227251c35d7bd1ed96) from the released weights of [OneJev1.6-0.8B](https://modelscope.cn/models/OmniJev/OneJev1.6-0.8B).
| File | Size | |
|---|---|---|
| `OneJev1.6-0.8B-Q4_K_M.gguf` | 529 MB | smallest, for low memory |
| `OneJev1.6-0.8B-Q5_K_M.gguf` | 578 MB | |
| `OneJev1.6-0.8B-Q6_K.gguf` | 630 MB | |
| `OneJev1.6-0.8B-Q8_0.gguf` | 812 MB | recommended, closest to the original |
| `OneJev1.6-0.8B-F16.gguf` | 1.5 GB | full precision |
| `mmproj-OneJev1.6-0.8B-F16.gguf` | 205 MB | vision projector, needed for images |
## Decision Index
On the public part of [Decision Index 0.3](https://huggingface.co/spaces/multimodalart/jev-decision-index) (37 benchmarks, 140,178 requests, run and scored with the board's own kit), OneJev1.6-0.8B scores 48.82, up from 42.71 for OneJev1.5-0.8B. That is the highest of any model up to 2.5B parameters on the board, and it leads its class in all five areas.
The four other models are the highest-scoring models of up to 2.5B parameters on the board.
It also scores above 47 models of 4B parameters and more on the board.
Other models' scores are their public-part scores on the board on 16 October 2026.
## 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.
## Quick start
Install [llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) first (`brew install llama.cpp` on macOS), then download one model file and the vision projector and serve them with qev.
```bash
pip install modelscope git+https://github.com/OmniJev/OneJev.git
modelscope download --model OmniJev/OneJev1.6-0.8B-GGUF OneJev1.6-0.8B-Q8_0.gguf mmproj-OneJev1.6-0.8B-F16.gguf --local_dir OneJev1.6-0.8B-GGUF
qev serve --gguf OneJev1.6-0.8B-GGUF/OneJev1.6-0.8B-Q8_0.gguf --port 8000
```
qev finds the vision projector next to the model file and fetches the tokenizer from Hugging Face. Without Hugging Face access, download the tokenizer from ModelScope and pass it with `--tokenizer`.
```bash
modelscope download --model OmniJev/OneJev1.6-0.8B --include "*.json" "*.jinja" --local_dir OneJev1.6-0.8B
qev serve --gguf OneJev1.6-0.8B-GGUF/OneJev1.6-0.8B-Q8_0.gguf --tokenizer OneJev1.6-0.8B --port 8000
```
```python
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": ""},
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. For video, use the PyTorch weights in [OneJev1.6-0.8B](https://modelscope.cn/models/OmniJev/OneJev1.6-0.8B). More examples and the code are on [GitHub](https://github.com/OmniJev/OneJev).
## Citation
```bibtex
@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
