Перейти к основному содержанию

Qwen3 VL 4B (Reasoning)

AlibabaQwenОткрытые весаApache 2.0 · Коммерческое использование

Описание

Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video. Built on a 235B-parameter architecture, it integrates early joint training of visual and textual modalities for strong language grounding. The model supports up to a 1 million-token context window and excels at visual understanding, spatial reasoning, long video comprehension, and tool-based interaction. It can generate code from images, perform precise 2D/3D object grounding, and operate digital interfaces like a visual agent. The “Instruct” version rivals Gemini 2.5 Pro in perception benchmarks, while the “Thinking” version leads in multimodal reasoning and STEM tasks. With multilingual OCR, creative writing, and fine-grained scene interpretation, Qwen3-VL establishes a new open-source frontier for integrated vision-language intelligence.

Дата выхода
2025-10-14
Параметры
4.0B
Длина контекста
—
Модальности
image, text

Радар способностей

26
general
32
coding
30
reasoning
36
science
70
agents
100
multimodal

Рейтинги

Домен#МестоОценкаИсточник
Рейтинг кодинга499
20.0
AA
Общий рейтинг492
28.0
AA
Мультимодальный рейтинг139
25.0
LS
Наука506
26.0
AA

Оценки бенчмарков (LLM Stats)

(LLM Stats (zeroeval))

Chat

MM-MT-Bench7.70 / 100Сам.
IFEvalGoogle Research (2023)82.6%Сам.
Multi-IF73.6%Сам.

Creativity

Creative Writing v376.1%Сам.

General

MMLU81.5%Сам.
MLVU-M75.7%Сам.
BFCL-v367.3%Сам.
Include64.6%Сам.
Arena-Hard v236.8%Сам.

Language

MMLU-Redux86.0%Сам.
MMLU-Pro73.6%Сам.
MMLU-ProX65.0%Сам.

Math

MathVista-Mini79.5%Сам.
AIME 202574.5%Сам.
LiveBench 2024112568.4%Сам.
MathVision60.0%Сам.
HMMT2553.1%Сам.
PolyMATH44.6%Сам.

Multimodal

VideoMMMU69.4%Сам.
OSWorld31.4%Сам.

Reasoning

CharXiv-D83.9%Сам.
GPQANYU + Cohere + Anthropic (2023)64.1%Сам.
LiveCodeBench v651.3%Сам.
CharXiv-R50.3%Сам.
SuperGPQA46.8%Сам.

Video

CharadesSTA59.0%Сам.

Vision

DocVQAtest94.2%Сам.
ScreenSpot92.9%Сам.
MMBench-V1.186.7%Сам.
AI2D84.9%Сам.
InfoVQAtest83.0%Сам.
OCRBench80.8%Сам.
MuirBench75.0%Сам.
CC-OCR73.8%Сам.
MMStar73.2%Сам.
RealWorldQA73.2%Сам.
MMMU (val)70.8%Сам.
MVBench69.3%Сам.
Hallusion Bench64.1%Сам.
BLINK63.4%Сам.
OCRBench-V2 (en)61.8%Сам.
MMMU-Pro57.0%Сам.
OCRBench-V2 (zh)55.8%Сам.
LVBench53.5%Сам.
ScreenSpot Pro49.2%Сам.
ERQA47.3%Сам.
ODinW39.4%Сам.

Writing

WritingBench84.0%Сам.

Индексы оценки AA

(Artificial Analysis)
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
70.0
Gpqa(NYU + Cohere + Anthropic (2023))
49.4
Ifbench(Google Research (2023))
36.6
Livecodebench(UC Berkeley + MIT + Cornell (2024))
32.0
Math Index(Artificial Analysis)
25.7
Aime 25(MAA (Mathematical Association of America))
25.7
Lcr(Artificial Analysis)
21.3
Tau2(Sierra + U Toronto + Vector Institute (2025))
15.5
Intelligence Index(Artificial Analysis)
7.0
Hle(Center for AI Safety + Scale AI (2025))
4.6
Terminalbench Hard(Stanford × Laude Institute (2026))
1.5

Оценки категорий LLM Stats

(LLM Stats (zeroeval))
Chat
3
Communication
3
Multimodal
100
Instruction Following
80
Structured Output
80
General
80
Image To Text
70
Language
70
Legal
70
Math
70
Reasoning
70
Finance
70
Grounding
70
Healthcare
70
Creativity
70
Text-to-image
70
Tool Calling
70
Vision
70
Writing
70
Physics
60
Spatial Reasoning
60
3d
60
Biology
60
Chemistry
60
Video
60
Long Context
50
Agents
50
Economics
50

Цены

Цена вводаБесплатно
Цена выводаБесплатно
Смешанная цена (3:1)Бесплатно

Скорость

Токенов/сек0.0
Задержка первого токена0.00s
Время до первого ответа0.00s

Рейтинг цен провайдеров

Нет данных провайдеров

Внешние ссылки