跳轉到主要內容

Qwen3 VL 4B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

描述

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.

發布日期
2025-10-14
參數規模
4.0B
上下文長度
1.0M
支援模態
image, text

能力雷達圖

29
general
17
coding
30
reasoning
29
science估算
70
agents
100
multimodal

Science 在缺少專門科學評測時使用推理能力代理估算。

排行榜排名

領域#排名分數來源
智能体与工具65
53.0
LS
代码能力榜370
17.0
AA
通用能力榜353
32.0
AA
数学推理285
26.0
AA
多模态榜51
74.0
LS
推理能力77
54.0
LS
科学能力378
28.0
AA

基準測試分數 (LLM Stats)

3d

BLINK63.4%自報

Agents

BFCL-v367.3%自報
OSWorld31.4%自報

Biology

GPQA64.1%自報

Chemistry

SuperGPQA46.8%自報

Communication

MM-MT-Bench7.70 / 100自報
WritingBench84.0%自報
Multi-IF73.6%自報

Creativity

Creative Writing v376.1%自報
Arena-Hard v236.8%自報

Finance

MMLU81.5%自報
MMLU-Pro73.6%自報
MMLU-ProX65.0%自報

General

MMLU-Redux86.0%自報
IFEval82.6%自報
MLVU-M75.7%自報
MMStar73.2%自報
MMMU (val)70.8%自報
LiveBench 2024112568.4%自報
Include64.6%自報
MMMU-Pro57.0%自報
LiveCodeBench v651.3%自報

Grounding

ScreenSpot92.9%自報
ScreenSpot Pro49.2%自報

Healthcare

VideoMMMU69.4%自報

Image To Text

OCRBench80.8%自報
OCRBench-V2 (en)61.8%自報
OCRBench-V2 (zh)55.8%自報

Language

CharadesSTA59.0%自報

Long Context

LVBench53.5%自報

Math

MathVista-Mini79.5%自報
AIME 202574.5%自報
MathVision60.0%自報
HMMT2553.1%自報
PolyMATH44.6%自報

Multimodal

DocVQAtest94.2%自報
MMBench-V1.186.7%自報
AI2D84.9%自報
CharXiv-D83.9%自報
InfoVQAtest83.0%自報
MuirBench75.0%自報
CC-OCR73.8%自報
MVBench69.3%自報
CharXiv-R50.3%自報

Reasoning

Hallusion Bench64.1%自報
ERQA47.3%自報

Spatial Reasoning

RealWorldQA73.2%自報

Vision

ODinW39.4%自報

AA 評測指數

Math Index
25.7
Intelligence Index
13.7
Coding Index
6.7
Mmlu Pro
0.7
Gpqa
0.5
Ifbench
0.4
Livecodebench
0.3
Aime 25
0.3
Lcr
0.2
Scicode
0.2
Tau2
0.2
Hle
0.0
Terminalbench Hard
0.0

LLM Stats 分類評分

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

定價

輸入價格免費
輸出價格免費
混合價格(3:1)免費

速度

Tokens/秒0.0 tokens/s
首Token延遲0.00s
首回答延遲0.00s

可用提供商

(LS 內部計價單位)
提供商輸入價格輸出價格
DeepInfra100K1.0M

外部連結