Unsloth Dynamic 2.0 实现了卓越的准确性,性能超越其他主流量化方案。
欢迎体验 Qwen3-VL——迄今为止 Qwen 系列中功能最为强大的视觉语言模型。
本代模型实现了全面升级:文本理解与生成能力更优、视觉感知与推理深度更强、上下文长度显著扩展、空间与视频动态理解能力提升,以及智能体交互能力增强。
提供 Dense 和 MoE 两种架构,可从边缘设备到云端灵活扩展;同时推出 Instruct 版本与推理增强的 Thinking 版本,满足按需部署需求。
视觉智能体:可操控电脑/手机图形界面——识别界面元素、理解功能、调用工具、完成任务。
视觉编码增强:能从图像/视频生成 Draw.io 图表及 HTML/CSS/JS 代码。
高级空间感知:判断物体位置、视角和遮挡关系;提供更强的 2D 定位能力,并支持 3D 定位,赋能空间推理与具身智能。
长上下文与视频理解:原生支持 256K 上下文,可扩展至 100 万;轻松处理整本书籍和长达数小时的视频,实现完整回忆与秒级索引。
增强多模态推理:在 STEM/数学领域表现出色——擅长因果分析,提供基于证据的逻辑答案。
升级视觉识别:通过更广泛、更高质量的预训练,实现“万物可识”——涵盖名人、动漫、产品、地标、动植物等。
扩展 OCR 功能:支持 32 种语言(此前为 19 种);在低光、模糊、倾斜场景下表现稳定;对生僻/古文字和专业术语识别更精准;长文档结构解析能力提升。
文本理解能力媲美纯语言模型:实现文本-视觉无缝融合,确保无损、统一的理解体验。
Interleaved-MRoPE:通过稳健的位置嵌入,在时间、宽度和高度维度实现全频率分配,提升长时视频推理能力。
DeepStack:融合多级别 ViT 特征,捕捉细粒度细节,增强图文对齐精度。
文本-时间戳对齐:超越 T-RoPE,实现精确的时间戳级事件定位,强化视频时序建模。
本仓库为 Qwen3-VL-4B-Instruct 的权重库。
多模态性能

纯文本性能

以下提供简单示例,展示如何结合🤖 ModelScope和🤗 Transformers使用Qwen3-VL。
Qwen3-VL的代码已集成到最新版Hugging Face transformers中,建议通过以下命令从源代码构建:
pip install git+https://github.com/huggingface/transformers
# pip install transformers==4.57.0 # currently, V4.57.0 is not released以下是一个代码片段,展示如何使用 transformers 调用对话模型:
from transformers import Qwen3VLForConditionalGeneration, AutoProcessor
# default: Load the model on the available device(s)
model = Qwen3VLForConditionalGeneration.from_pretrained(
"Qwen/Qwen3-VL-4B-Instruct", dtype="auto", device_map="auto"
)
# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
# model = Qwen3VLForConditionalGeneration.from_pretrained(
# "Qwen/Qwen3-VL-4B-Instruct",
# dtype=torch.bfloat16,
# attn_implementation="flash_attention_2",
# device_map="auto",
# )
processor = AutoProcessor.from_pretrained("Qwen/Qwen/Qwen3-VL-4B-Instruct")
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
},
{"type": "text", "text": "Describe this image."},
],
}
]
# Preparation for inference
inputs = processor.apply_chat_template(
messages,
tokenize=True,
add_generation_prompt=True,
return_dict=True,
return_tensors="pt"
)
# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)如果您发现我们的工作对您有所帮助,欢迎引用我们的成果。
@misc{qwen3technicalreport,
title={Qwen3 Technical Report},
author={Qwen Team},
year={2025},
eprint={2505.09388},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2505.09388},
}
@article{Qwen2.5-VL,
title={Qwen2.5-VL Technical Report},
author={Bai, Shuai and Chen, Keqin and Liu, Xuejing and Wang, Jialin and Ge, Wenbin and Song, Sibo and Dang, Kai and Wang, Peng and Wang, Shijie and Tang, Jun and Zhong, Humen and Zhu, Yuanzhi and Yang, Mingkun and Li, Zhaohai and Wan, Jianqiang and Wang, Pengfei and Ding, Wei and Fu, Zheren and Xu, Yiheng and Ye, Jiabo and Zhang, Xi and Xie, Tianbao and Cheng, Zesen and Zhang, Hang and Yang, Zhibo and Xu, Haiyang and Lin, Junyang},
journal={arXiv preprint arXiv:2502.13923},
year={2025}
}
@article{Qwen2VL,
title={Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution},
author={Wang, Peng and Bai, Shuai and Tan, Sinan and Wang, Shijie and Fan, Zhihao and Bai, Jinze and Chen, Keqin and Liu, Xuejing and Wang, Jialin and Ge, Wenbin and Fan, Yang and Dang, Kai and Du, Mengfei and Ren, Xuancheng and Men, Rui and Liu, Dayiheng and Zhou, Chang and Zhou, Jingren and Lin, Junyang},
journal={arXiv preprint arXiv:2409.12191},
year={2024}
}
@article{Qwen-VL,
title={Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond},
author={Bai, Jinze and Bai, Shuai and Yang, Shusheng and Wang, Shijie and Tan, Sinan and Wang, Peng and Lin, Junyang and Zhou, Chang and Zhou, Jingren},
journal={arXiv preprint arXiv:2308.12966},
year={2023}
}