跳轉到主要內容

Qwen3 VL 30B A3B Instruct

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
2025-10-03
參數規模
31.0B
上下文長度
131K
支援模態
image, text, video

能力雷達圖

33
general
28
coding
72
reasoning
43
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69
52.0
LS
代码能力榜278
28.0
AA
通用能力榜318
35.0
AA
数学推理110
73.0
AA
多模态榜45
75.0
LS
推理能力85
51.0
LS
科学能力234
45.0
AA

基準測試分數 (LLM Stats)

3d

BLINK67.7%自報

Agents

BFCL-v366.3%自報
OSWorld30.3%自報

Biology

GPQA70.4%自報

Chemistry

SuperGPQA53.1%自報

Communication

MM-MT-Bench8.10 / 100自報
WritingBench82.6%自報
Multi-IF66.1%自報

Creativity

Creative Writing v384.6%自報
Arena-Hard v258.5%自報

Factuality

SimpleQA27.0%自報

Finance

MMLU85.0%自報
MMLU-Pro77.8%自報
MMLU-ProX70.9%自報

General

MMLU-Redux88.4%自報
IFEval85.8%自報
MLVU-M81.3%自報
MMMU (val)74.2%自報
MMStar72.1%自報
Include71.6%自報
LiveBench 2024112565.4%自報
MMMU-Pro60.4%自報
LiveCodeBench v642.6%自報

Grounding

ScreenSpot94.7%自報
ScreenSpot Pro60.5%自報

Healthcare

VideoMMMU68.7%自報

Image To Text

OCRBench90.3%自報
OCRBench-V2 (en)63.2%自報
OCRBench-V2 (zh)57.8%自報

Language

CharadesSTA63.5%自報

Long Context

LVBench62.5%自報

Math

MathVista-Mini80.1%自報
AIME 202569.3%自報
MathVision60.2%自報
HMMT2550.6%自報
PolyMATH44.3%自報

Multimodal

DocVQAtest95.0%自報
MMBench-V1.187.0%自報
CharXiv-D85.5%自報
AI2D85.0%自報
InfoVQAtest82.0%自報
CC-OCR80.7%自報
Video-MME74.5%自報
MVBench72.3%自報
MuirBench62.9%自報
CharXiv-R48.9%自報

Reasoning

Hallusion Bench61.5%自報
ERQA43.0%自報

Spatial Reasoning

RealWorldQA73.7%自報

Vision

ODinW47.5%自報

AA 評測指數

Math Index
72.3
Intelligence Index
16.1
Coding Index
14.3
Mmlu Pro
0.8
Aime 25
0.7
Gpqa
0.7
Livecodebench
0.5
Ifbench
0.3
Scicode
0.3
Lcr
0.2
Tau2
0.2
Hle
0.1
Terminalbench Hard
0.1

LLM Stats 分類評分

Communication
3
Multimodal
100
Structured Output
80
Text-to-image
80
Writing
80
Creativity
80
Grounding
80
Instruction Following
80
Language
80
Spatial Reasoning
70
Tool Calling
70
Video
70
Vision
70
3d
70
Biology
70
Finance
70
General
70
Healthcare
70
Image To Text
70
Legal
70
Math
70
Reasoning
70
Chemistry
60
Long Context
60
Physics
60
Agents
50
Economics
50
Factuality
30

定價

輸入價格$0.2 / 1M tokens
輸出價格$0.6 / 1M tokens
混合價格(3:1)$0.3 / 1M tokens

速度

Tokens/秒125.3 tokens/s
首Token延遲1.08s
首回答延遲1.08s

可用提供商

(LS 內部計價單位)
提供商輸入價格輸出價格
Novita200K700K
DeepInfra290K990K

外部連結