我们很高兴地宣布对 MiMo-VL 模型(MiMo-VL-7B-RL-2508 和 MiMo-VL-7B-SFT-2508)进行了改进,在多个基准测试中性能得到提升,思维控制能力增强,用户体验也更为出色。
MiMo-VL-7B-RL-2508 在图像和视频基准测试中均表现出持续改进,取得了 MMMU 70.6 分和 VideoMME 70.8 分的显著成绩。
完整评估结果可参见下文。
新增思维控制功能,用户可通过 no_think 参数关闭模型的推理模式:
内部 VLM Arena 评分显示,模型在实际场景中的表现有显著提升:
这些更新使视觉语言模型在学术评估和实际应用中更具能力、更灵活且更可靠。

MiMo-VL-7B-RL-2508 的回答:
To determine the appealing features of the Xiaomi SU7, we analyze visual elements and context from the video:
1. Aesthetic Design
- The car boasts a sleek, aerodynamic exterior with smooth, modern lines.
- Distinctive lighting (sharp headlights and a continuous, glowing taillight strip) enhances its futuristic and sporty look.
- Vibrant color and polished finish make it visually striking.
2. Performance & Speed
- Multiple dynamic shots of the car driving at high speeds on diverse roads (desert, forest, city) emphasize agility and power.
- Close - ups of spinning wheels and the engine (with lightning-like effects) suggest advanced engineering and high-performance capabilities (likely electric or tech-driven).
3. Versatility
- The car is shown in varied environments (desert at dusk, forested roads, urban streets), highlighting its adaptability to different terrains and use cases.
4. Technological Innovation
- The engine close-up with electric-like lightning effects implies cutting-edge technology, positioning the car as a tech-forward vehicle (aligning with Xiaomi’s tech brand identity).
5. Brand Appeal
Prominent “Xiaomi SU7” branding leverages Xiaomi’s reputation in consumer tech, adding credibility and allure for tech-savvy audiences.
In summary, the Xiaomi SU7 appeals through its striking design, high-performance capabilities, versatility, technological innovation, and strong brand association.
MiMo-VL-7B-2508 模型的两个版本现已在 Hugging Face 开源:
chat_template.json 中设置,无需额外配置。用户可通过在查询后附加 /no_think 来控制思维模式:
❗️重要提示:/no_think 命令必须是用户消息的最后部分,这意味着 /no_think 之后不应有任何用户内容,如图像或视频。
对于包含单张图像或视频的提示,务必将视觉媒体放在文本之前。例如:
✅ 正确示例:
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": image_path},
{"type": "text", "text": "Describe the image. /no_think"},
],
}
]❌ 不当示例:
messages = [
{
"role": "user",
"content": [
{"type": "text", "text": "Describe the image. /no_think"},
{"type": "image", "image": image_path},
],
}
]本报告将分享我们构建轻量且高性能视觉语言模型(VLM)MiMo-VL-7B的研究成果。MiMo-VL-7B 由三部分组成:(1) 原生分辨率 ViT 编码器,用于保留细粒度视觉细节;(2) MLP 投影器,实现高效跨模态对齐;(3) 我们的 MiMo-7B 语言模型,该模型针对复杂推理任务进行了专门优化。
MiMo-VL-7B 的开发分为两个连续的训练阶段:(1) 四阶段预训练阶段,包括投影器预热、视觉-语言对齐、通用多模态预训练和长上下文监督微调(SFT),最终得到 MiMo-VL-7B-SFT 模型。(2) 后续的后训练阶段,我们引入了混合在线强化学习(MORL)框架——这是一种新颖的框架,可无缝整合涵盖感知准确性、视觉定位精度、逻辑推理能力以及人类/AI 偏好的多种奖励信号,最终得到 MiMo-VL-7B-RL 模型。
我们开源了 MiMo-VL-7B 系列模型,包括 SFT 和 RL 模型的检查点。我们相信本报告及所发布的模型将为开发高性能推理型 VLM 提供有价值的见解,并造福更广泛的社区。
模型可在 Huggingface Collections: MiMo-VL 和 ModelScope Collections: MiMo-VL 获取
| 模型 | 描述 | 下载(HuggingFace) | 下载(ModelScope) |
|---|---|---|---|
| MiMo-VL-7B-SFT | 经过四阶段预训练后具备卓越推理潜力的视觉语言模型 | 🤗 XiaomiMiMo/MiMo-VL-7B-SFT | 🤖️ XiaomiMiMo/MiMo-VL-7B-SFT |
| MiMo-VL-7B-RL | 超越现有开源模型的强化学习模型 | 🤗 XiaomiMiMo/MiMo-VL-7B-RL | 🤖️ XiaomiMiMo/MiMo-VL-7B-RL |
在通用视觉语言理解方面,MiMo-VL-7B 模型取得了开源领域的最先进成果。
在多模态推理任务中,SFT 和 RL 模型在所有对比基准测试中均显著优于其他开源基线模型。
[!IMPORTANT] 标有 * 的结果通过我们的评估框架获得。 标有 的任务由 GPT-4o 进行评估。
MiMo-VL-7B-RL 具备卓越的 GUI 理解与定位能力。作为一款通用型视觉语言模型,MiMo-VL 的性能已达到甚至超越了 GUI 专用模型。
依托自研评估数据集及 GPT-4o 评判体系,MiMo-VL-7B-RL 在所有参评开源视觉语言模型中获得最高 Elo 评分,在参数规模从 70 亿到 720 亿的模型中均排名第一。
MiMo-VL-7B 系列在部署与推理方面完全兼容 Qwen2_5_VLForConditionalGeneration 架构。
@misc{coreteam2025mimovltechnicalreport,
title={MiMo-VL Technical Report},
author={LLM-Core-Team Xiaomi},
year={2025},
eprint={2506.03569},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2506.03569},
}如有任何问题,请通过 mimo@xiaomi.com 与我们联系,或提交 issue。