跳转到主要内容

Qwen2.5 VL 7B Instruct

Alibaba Cloud / Qwen TeamQwenOpen WeightApache 2.0 · Commercial OK

描述

Qwen2.5-VL is a vision-language model from the Qwen family. Key enhancements include visual understanding (objects, text, charts, layouts), visual agent capabilities (tool use, computer/phone control), long video comprehension with event pinpointing, visual localization (bounding boxes/points), and structured output generation.

发布日期
2025-01-26
参数规模
8.3B
上下文长度
支持模态

能力雷达图

50
general
0
coding
50
reasoning
51
science估算
50
agents
90
multimodal

Science 在缺少专门科学评测时使用推理能力代理估算。

排行榜排名

领域#排名分数来源
智能体与工具35
62.0
LS
多模态榜59
71.0
LS
推理能力82
53.0
LS

基准测试分数 (LLM Stats)

Agents

MobileMiniWob++_SR91.4%自报
AITZ_EM81.9%自报
AndroidWorld_SR25.5%自报

General

MMVet67.1%自报
MMStar63.9%自报
MMT-Bench63.6%自报
MMMU58.6%自报
MMMU-Pro38.3%自报

Grounding

ScreenSpot84.7%自报
ScreenSpot Pro29.0%自报

Image To Text

DocVQA95.7%自报
OCRBench86.4%自报
TextVQA84.9%自报

Language

CharadesSTA43.6%自报

Long Context

MLVU70.2%自报
LongVideoBench54.7%自报
LVBench45.3%自报

Math

MathVista-Mini68.2%自报
MathVision25.1%自报

Multimodal

Android Control Low_EM91.4%自报
ChartQA87.3%自报
MMBench84.3%自报
InfoVQA82.6%自报
CC-OCR77.8%自报
TempCompass71.7%自报
VideoMME w sub.71.6%自报
PerceptionTest70.5%自报
MVBench69.6%自报
VideoMME w/o sub.65.1%自报
Android Control High_EM60.1%自报
MMBench-Video1.8%自报

Reasoning

Hallusion Bench52.9%自报

AA 评测指数

暂无 AA 评测数据

LLM Stats 分类评分

Image To Text
90
Structured Output
80
Text-to-image
80
Spatial Reasoning
60
Vision
60
Grounding
60
Healthcare
60
Long Context
60
Multimodal
60
Reasoning
60
Video
50
Agents
50
General
50
Math
50
Language
40

定价

暂无定价数据

速度

暂无速度数据

可用提供商

(LS 内部计价单位)

暂无提供商数据

外部链接