跳转到主要内容

Qwen3 VL 30B A3B (Reasoning)

AlibabaQwen开源权重Apache 2.0 · 商用许可

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.

发布日期
2025-10-03
参数规模
31.0B
上下文长度
131K
支持模态
image, text, video

能力雷达图

31
general
70
coding
81
reasoning
53
science
70
agents
100
multimodal

排行榜排名

领域#排名分数来源
代码能力榜358
42.0
AA
通用能力榜373
36.0
AA
多模态榜113
41.0
LS
科学能力339
43.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Chat

MM-MT-Bench7.90 / 100自报
IFEvalGoogle Research (2023)81.7%自报
Multi-IF73.0%自报

Creativity

Creative Writing v382.5%自报

Factuality

SimpleQA23.9%自报

General

MMLU87.6%自报
MLVU-M78.9%自报
Include74.5%自报
BFCL-v368.6%自报
Arena-Hard v256.7%自报

Language

MMLU-Redux90.9%自报
MMLU-Pro80.5%自报
MMLU-ProX76.1%自报

Math

AIME 202583.1%自报
MathVista-Mini81.9%自报
LiveBench 2024112572.1%自报
HMMT2567.6%自报
MathVision65.7%自报
PolyMATH51.7%自报

Multimodal

VideoMMMU75.0%自报
Video-MME73.3%自报
OSWorld30.6%自报

Reasoning

CharXiv-D86.9%自报
GPQANYU + Cohere + Anthropic (2023)74.4%自报
LiveCodeBench v664.2%自报
CharXiv-R56.6%自报
SuperGPQA56.4%自报

Video

CharadesSTA62.7%自报

Vision

DocVQAtest95.0%自报
ScreenSpot94.7%自报
MMBench-V1.188.9%自报
AI2D86.9%自报
InfoVQAtest86.0%自报
OCRBench83.9%自报
CC-OCR77.8%自报
MuirBench77.6%自报
RealWorldQA77.4%自报
MMMU (val)76.0%自报
MMStar75.5%自报
MVBench72.0%自报
Hallusion Bench66.0%自报
BLINK65.4%自报
MMMU-Pro63.0%自报
OCRBench-V2 (en)62.6%自报
OCRBench-V2 (zh)60.4%自报
LVBench59.2%自报
ScreenSpot Pro57.3%自报
ERQA45.3%自报
ODinW42.3%自报

Writing

WritingBench85.2%自报

AA 评测指数

(Artificial Analysis)
Aime 25(MAA (Mathematical Association of America))
82.3
Math Index(Artificial Analysis)
82.3
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
80.7
Gpqa(NYU + Cohere + Anthropic (2023))
72.0
Livecodebench(UC Berkeley + MIT + Cornell (2024))
69.7
Ifbench(Google Research (2023))
45.1
Tau2(Sierra + U Toronto + Vector Institute (2025))
19.9
Intelligence Index(Artificial Analysis)
9.5
Hle(Center for AI Safety + Scale AI (2025))
8.9
Terminalbench Hard(Stanford × Laude Institute (2026))
5.3

LLM Stats 分类评分

(LLM Stats (zeroeval))
Communication
3
Chat
3
Multimodal
100
General
90
Instruction Following
80
Language
80
Legal
80
Structured Output
80
Finance
80
Grounding
80
Healthcare
80
Text-to-image
80
Image To Text
70
Math
70
Physics
70
Reasoning
70
Spatial Reasoning
70
3d
70
Biology
70
Chemistry
70
Creativity
70
Tool Calling
70
Video
70
Vision
70
Writing
70
Long Context
60
Economics
60
Agents
50
Factuality
20

定价

输入价格$0.2 / 1M tokens
输出价格$2.4 / 1M tokens
混合价格(3:1)$0.75 / 1M tokens

速度

Tokens/秒0.0
首Token延迟0.00s
首回答延迟0.00s

供应商价格排行

供应商价格排行

5 个供应商

最便宜: Alibaba (China)最贵: Kilo Gateway
供应商输入输出
1Alibaba (China)最便宜
$0.108
$0.431
2Alibaba主要
$0.2
$2.4
3OpenRouter
$0.2
$2.4
4NovitaAI
$0.2
$1
5Kilo Gateway
$0.2
$2.4

比较该模型在不同 API 供应商之间的定价。

外部链接