Qwen2.5 VL 72B Instruct
Alibaba Cloud / Qwen TeamQwenOpen Weighttongyi-qianwen
Описание
Qwen2.5-VL is the new flagship vision-language model of Qwen, significantly improved from Qwen2-VL. It excels at recognizing objects, analyzing text/charts/layouts in images, acting as a visual agent, understanding long videos (over 1 hour) with event pinpointing, performing visual localization (bounding boxes/points), and generating structured outputs from documents.
Дата выхода
2025-01-26
Параметры
72.0B
Длина контекста
32K
Модальности
image, text
Радар способностей
50
general
0
coding
60
reasoning
60
scienceоцен.
40
agents
80
multimodal
Science использует прокси на основе рассуждений, когда специализированные научные бенчмарки недоступны.
Рейтинги
| Домен | #Место | Оценка | Источник |
|---|---|---|---|
| Agents & Tools | 82 | 45.0 | LS |
| Multimodal Ranking | 54 | 73.0 | LS |
| Reasoning | 74 | 55.0 | LS |
Оценки бенчмарков (LLM Stats)
Agents
AITZ_EM
83.2%Сам.
MobileMiniWob++_SR
68.0%Сам.
AndroidWorld_SR
35.0%Сам.
OSWorld
8.8%Сам.
General
MMVet
76.2%Сам.
MLVU-M
74.6%Сам.
MMStar
70.8%Сам.
MMMU
70.2%Сам.
MMMU-Pro
51.1%Сам.
Grounding
ScreenSpot
87.1%Сам.
ScreenSpot Pro
43.6%Сам.
Image To Text
DocVQA
96.4%Сам.
OCRBench
88.5%Сам.
OCRBench-V2 (en)
61.5%Сам.
Long Context
EgoSchema
76.2%Сам.
LVBench
47.3%Сам.
Math
MathVista-Mini
74.8%Сам.
MathVision
38.1%Сам.
Multimodal
Android Control Low_EM
93.7%Сам.
ChartQA
89.5%Сам.
AI2D
88.4%Сам.
MMBench
88.0%Сам.
CC-OCR
79.8%Сам.
TempCompass
74.8%Сам.
VideoMME w/o sub.
73.3%Сам.
PerceptionTest
73.2%Сам.
MVBench
70.4%Сам.
Android Control High_EM
67.4%Сам.
MMBench-Video
2.0%Сам.
Reasoning
Hallusion Bench
55.2%Сам.
Индексы оценки AA
Нет данных AA оценки
Оценки категорий LLM Stats
Structured Output80
Text-to-image80
Image To Text80
Spatial Reasoning70
Grounding70
Healthcare70
Reasoning70
Vision60
Long Context60
Math60
Multimodal60
Video50
General50
Agents40
Цены
Цена ввода$0.25 / 1M tokens
Цена вывода$0.75 / 1M tokens
Смешанная цена (3:1)$0.375 / 1M tokens
Скорость
Нет данных о скорости
Доступные провайдеры
(Внутренние единицы LS)Нет данных провайдеров