---
title: ScaleCUA-glm4.6v-osworld
canonical_url: "https://www.modelscope.cn/models/extreme1228/ScaleCUA-glm4.6v-osworld"
md_url: "https://www.modelscope.cn/models/extreme1228/ScaleCUA-glm4.6v-osworld.md"
repository: extreme1228/ScaleCUA-glm4.6v-osworld
last_updated: 2026-07-09
license: other
pipeline_tag: image-text-to-text
tasks:
  - image-text-to-text
model_type:
  - glm4v
architectures:
  - Glm4vForConditionalGeneration
base_model:
  - extreme1228/ScaleCUA-glm4.6v-osworld-sft
base_model_relation: finetune
parameters: 10.3B
tensor_type:
  - BF16
library_name:
  - safetensors
downloads: 12
stars: 0
tags:
  - computer-use-agent
  - gui-agent
  - osworld
  - scalecua
---

# ScaleCUA-glm4.6v-osworld

> ScaleCUA-glm4.6v-osworld - extreme1228 在 ModelScope 开源的模型。ScaleCUA-glm4.6v-osworld

extreme1228/ScaleCUA-glm4.6v-osworld 是 ModelScope 魔搭社区上的 10.3B 参数image-text-to-text模型，采用 other 许可，基于 extreme1228/ScaleCUA-glm4.6v-osworld-sft 构建。

- **Repository**: extreme1228/ScaleCUA-glm4.6v-osworld
- **License**: other
- **Tasks**: image-text-to-text
- **Parameters**: 10.3B
- **Base model**: extreme1228/ScaleCUA-glm4.6v-osworld-sft
- **Tags**: computer-use-agent, gui-agent, osworld, scalecua
- **Downloads**: 12
- **Stars**: 0
- **Last updated**: 2026-07-09

Source: https://www.modelscope.cn/models/extreme1228/ScaleCUA-glm4.6v-osworld

---

# ScaleCUA-glm4.6v-osworld

A **GLM-4.6V-Flash** computer-use agent from **ScaleCUA**, trained for **OSWorld**.
Stage: **online-RL (RLVR)**.

- 📄 Paper / code: [https://github.com/THUDM/SCALE-CUA](https://github.com/THUDM/SCALE-CUA)
- 🤗 Collection: [https://huggingface.co/collections/extreme1228/scalecua-6a4cad5e3c80a230622eaff3](https://huggingface.co/collections/extreme1228/scalecua-6a4cad5e3c80a230622eaff3) · 🤖 [ModelScope](https://modelscope.cn/collections/extreme1228/ScaleCUA)

## What is this

This is the **online-RL checkpoint** (the released agent). Its SFT start point is [`extreme1228/ScaleCUA-glm4.6v-osworld-sft`](https://huggingface.co/extreme1228/ScaleCUA-glm4.6v-osworld-sft). It reaches **66.5%** on OSWorld.

ScaleCUA scales online RL for computer-use agents by pairing **VeriGen**
(verifiable GUI-task synthesis) with **Frontier Sampling** and **Visual Context
Segmentation**. See the repo for details.

## Usage

Serve with an OpenAI-compatible server (e.g. vLLM) and drive it with the
evaluation runtime in [`osworld_eval`](https://github.com/THUDM/SCALE-CUA/tree/main/osworld_eval):

```bash
pip install -U vllm
python -m vllm.entrypoints.openai.api_server --model extreme1228/ScaleCUA-glm4.6v-osworld \
  --served-model-name scalecua --trust-remote-code --port 8000
```

Then follow [`REPRODUCE.md`](https://github.com/THUDM/SCALE-CUA/blob/main/REPRODUCE.md) to run OSWorld evaluation.

## Results

**66.5%** on OSWorld (see the paper for the full table).

## License

Weights are released for research; the model inherits the license of its base
model (**GLM-4.6V-Flash**). Repository code is under the licenses in the GitHub repo.

## Citation

```bibtex
@misc{lv2026scalecua,
  title  = {ScaleCUA: Scaling Computer Use Agents with Verifiable Task Synthesis and Efficient Online RL},
  author = {Bowen Lv and Xiao Liu and Yanyu Ren and Hanyu Lai and Bohao Jing and Hanchen Zhang and Yanxiao Zhao and Shuntian Yao and Jie Tang and Yuxiao Dong},
  year   = {2026},
  url    = {https://github.com/THUDM/SCALE-CUA}
}
```