Kimi-VL 技术报告
摘要
我们推出 Kimi-VL,一款高效的开源混合专家(MoE)视觉语言模型(VLM),具备先进的多模态推理、长上下文理解以及强大的智能体能力——其语言解码器仅激活 28 亿参数(Kimi-VL-A3B)。Kimi-VL 在多个具有挑战性的领域表现出色:作为通用 VLM,Kimi-VL 在多轮智能体任务(例如 OSWorld)中表现优异,可与旗舰模型相媲美。此外,它在多种高难度视觉语言任务中也展现出卓越能力,包括大学级别的图像与视频理解、OCR、数学推理以及多图像理解。在对比评测中,它能有效与 GPT-4o-mini、Qwen2.5-VL-7B 和 Gemma-3-12B-IT 等前沿高效 VLM 竞争,并在多个关键领域超越 GPT-4o。Kimi-VL 还在长上下文处理与清晰感知方面取得进展。凭借 128K 扩展上下文窗口,Kimi-VL 可处理多种长输入,在 LongVideoBench 上取得 64.5 分,在 MMLongBench-Doc 上取得 35.1 分。其原生分辨率视觉编码器 MoonViT 进一步使其能够观察并理解超高分辨率的视觉输入,在 InfoVQA 上取得 83.2 分,在 ScreenSpot-Pro 上取得 34.5 分,同时在常规任务中保持较低的计算开销。在 Kimi-VL 的基础上,我们推出其进阶的长思考版本:Kimi-VL-Thinking-2506。该最新模型通过长链式思维(CoT)监督微调(SFT)与强化学习(RL)训练而成,具备强大的长程推理能力(在 MMMU 上 64.0、MMMU-Pro 上 46.3、MathVision 上 56.9、MathVista 上 80.1、VideoMMMU 上 65.2),同时保持稳健的通用能力。代码与模型已公开发布于 https://github.com/MoonshotAI/Kimi-VL。
引用
@article{arxiv.2504.07491,
title = {Kimi-VL Technical Report},
author = {Kimi Team and Angang Du and Bohong Yin and Bowei Xing and Bowen Qu and Bowen Wang and Cheng Chen and Chenlin Zhang and Chenzhuang Du and Chu Wei and Congcong Wang and Dehao Zhang and Dikang Du and Dongliang Wang and Enming Yuan and Enzhe Lu and Fang Li and Flood Sung and Guangda Wei and Guokun Lai and Han Zhu and Hao Ding and Hao Hu and Hao Yang and Hao Zhang and Haoning Wu and Haotian Yao and Haoyu Lu and Heng Wang and Hongcheng Gao and Huabin Zheng and Jiaming Li and Jianlin Su and Jianzhou Wang and Jiaqi Deng and Jiezhong Qiu and Jin Xie and Jinhong Wang and Jingyuan Liu and Junjie Yan and Kun Ouyang and Liang Chen and Lin Sui and Longhui Yu and Mengfan Dong and Mengnan Dong and Nuo Xu and Pengyu Cheng and Qizheng Gu and Runjie Zhou and Shaowei Liu and Sihan Cao and Tao Yu and Tianhui Song and Tongtong Bai and Wei Song and Weiran He and Weixiao Huang and Weixin Xu and Xiaokun Yuan and Xingcheng Yao and Xingzhe Wu and Xinhao Li and Xinxing Zu and Xinyu Zhou and Xinyuan Wang and Y. Charles and Yan Zhong and Yang Li and Yangyang Hu and Yanru Chen and Yejie Wang and Yibo Liu and Yibo Miao and Yidao Qin and Yimin Chen and Yiping Bao and Yiqin Wang and Yongsheng Kang and Yuanxin Liu and Yuhao Dong and Yulun Du and Yuxin Wu and Yuzhi Wang and Yuzi Yan and Zaida Zhou and Zhaowei Li and Zhejun Jiang and Zheng Zhang and Zhilin Yang and Zhiqi Huang and Zihao Huang and Zijia Zhao and Ziwei Chen and Zongyu Lin},
journal= {arXiv preprint arXiv:2504.07491},
year = {2025}
}
备注
Updated Kimi-VL-A3B-Thinking-2506 information