---
title: VideoTemp-o3
canonical_url: "https://www.modelscope.ai/models/Kwai-Keye/VideoTemp-o3"
md_url: "https://www.modelscope.ai/models/Kwai-Keye/VideoTemp-o3.md"
repository: Kwai-Keye/VideoTemp-o3
last_updated: 2026-05-29
license: mit
model_type:
  - qwen2_5_vl
architectures:
  - Qwen2_5_VLForConditionalGeneration
base_model:
  - Qwen/Qwen2.5-VL-7B-Instruct
base_model_relation: finetune
parameters: 8.3B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 160
stars: 0
---

# VideoTemp-o3

> VideoTemp-o3 - An open-source model by Kwai-Keye on ModelScope. VideoTemp-o3: Harmonizing Temporal Grounding and Video Understanding in Agentic Thinking-with-Videos

Kwai-Keye/VideoTemp-o3 is a 8.3B-parameter machine learning model on ModelScope. licensed under mit. derived from Qwen/Qwen2.5-VL-7B-Instruct.

- **Repository**: Kwai-Keye/VideoTemp-o3
- **License**: mit
- **Parameters**: 8.3B
- **Base model**: Qwen/Qwen2.5-VL-7B-Instruct
- **Downloads**: 160
- **Stars**: 0
- **Last updated**: 2026-05-29

Source: https://www.modelscope.ai/models/Kwai-Keye/VideoTemp-o3

---

<h2 align="center"> <b>VideoTemp-o3: Harmonizing Temporal Grounding and Video Understanding in Agentic Thinking-with-Videos</b></h2>
  
<div align="center" style="font-size: 15pt">

<a href='https://liuwq-bit.github.io/VideoTemp-o3'><img src='https://img.shields.io/badge/Project-Page-green'></a>
<a href='https://arxiv.org/abs/2602.07801'><img src='https://img.shields.io/badge/Arxiv-2602.07801-red'></a>
<a href='https://github.com/Kwai-Keye/VideoTemp-o3'><img src='https://img.shields.io/badge/Code-Github-blue?logo=github'></a>
<br>
<a href='https://huggingface.co/Kwai-Keye/VideoTemp-o3'><img src='https://img.shields.io/badge/Model-VideoTemp o3-orange'></a>
<a href='https://huggingface.co/datasets/Kwai-Keye/VideoTemp-o3'><img src='https://img.shields.io/badge/Dataset-SFT & RL-yellow'></a>
<a href='https://huggingface.co/datasets/Kwai-Keye/VideoTemp-Bench'><img src='https://img.shields.io/badge/Benchmark-VideoTemp Bench-blue'></a>

</div>

![image](https://cdn-uploads.huggingface.co/production/uploads/65a6797f838b3acc5358f583/qAeCGmvo-IjTIhifmEt7x.png)

Illustration of the agentic pipeline in VideoTemp-o3. Given a video QA pair, the model performs on-demand temporal grounding to locate the most relevant segment, then refines it iteratively. Finally, it produces a reliable answer grounded in the pertinent visual evidence.

## Citation

If you find our work useful, please consider citing:

```bibtex
@article{liu2026videotemp,
  title={VideoTemp-o3: Harmonizing Temporal Grounding and Video Understanding in Agentic Thinking-with-Videos},
  author={Liu, Wenqi and Wang, Yunxiao and Ma, Shijie and Liu, Meng and Su, Qile and Zhang, Tianke and Fan, Haonan and Liu, Changyi and Jiang, Kaiyu and Chen, Jiankang and Tang, Kaiyu and Wen, Bin and Yang, Fan and Gao, Tingting and Li, Han and Wei, Yinwei and Song, Xuemeng},
  journal={arXiv preprint arXiv:2602.07801},
  year={2026}
}
```