メインコンテンツへスキップ

Qwen3 VL 30B A3B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

説明

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.

リリース日
2025-10-03
パラメータ
31.0B
コンテキスト長
131K
モダリティ
image, text, video

能力レーダー

36
general
34
coding
81
reasoning
44
science推定
70
agents
100
multimodal

専門的な科学ベンチマークが利用できない場合、Scienceは推論プロキシを使用して推定します。

ランキング

ドメイン#順位スコアソース
Agents & Tools62
54.0
LS
Code Ranking233
37.0
AA
General Ranking251
42.0
AA
Math Reasoning75
83.0
AA
Multimodal Ranking38
77.0
LS
Reasoning80
53.0
LS
Science212
47.0
AA

ベンチマークスコア (LLM Stats)

3d

BLINK65.4%自己申告

Agents

BFCL-v368.6%自己申告
OSWorld30.6%自己申告

Biology

GPQA74.4%自己申告

Chemistry

SuperGPQA56.4%自己申告

Communication

MM-MT-Bench7.90 / 100自己申告
WritingBench85.2%自己申告
Multi-IF73.0%自己申告

Creativity

Creative Writing v382.5%自己申告
Arena-Hard v256.7%自己申告

Factuality

SimpleQA23.9%自己申告

Finance

MMLU87.6%自己申告
MMLU-Pro80.5%自己申告
MMLU-ProX76.1%自己申告

General

MMLU-Redux90.9%自己申告
IFEval81.7%自己申告
MLVU-M78.9%自己申告
MMMU (val)76.0%自己申告
MMStar75.5%自己申告
Include74.5%自己申告
LiveBench 2024112572.1%自己申告
LiveCodeBench v664.2%自己申告
MMMU-Pro63.0%自己申告

Grounding

ScreenSpot94.7%自己申告
ScreenSpot Pro57.3%自己申告

Healthcare

VideoMMMU75.0%自己申告

Image To Text

OCRBench83.9%自己申告
OCRBench-V2 (en)62.6%自己申告
OCRBench-V2 (zh)60.4%自己申告

Language

CharadesSTA62.7%自己申告

Long Context

LVBench59.2%自己申告

Math

AIME 202583.1%自己申告
MathVista-Mini81.9%自己申告
HMMT2567.6%自己申告
MathVision65.7%自己申告
PolyMATH51.7%自己申告

Multimodal

DocVQAtest95.0%自己申告
MMBench-V1.188.9%自己申告
CharXiv-D86.9%自己申告
AI2D86.9%自己申告
InfoVQAtest86.0%自己申告
CC-OCR77.8%自己申告
MuirBench77.6%自己申告
Video-MME73.3%自己申告
MVBench72.0%自己申告
CharXiv-R56.6%自己申告

Reasoning

Hallusion Bench66.0%自己申告
ERQA45.3%自己申告

Spatial Reasoning

RealWorldQA77.4%自己申告

Vision

ODinW42.3%自己申告

AA評価指数

Math Index
82.3
Intelligence Index
19.7
Coding Index
13.1
Aime 25
0.8
Mmlu Pro
0.8
Gpqa
0.7
Livecodebench
0.7
Ifbench
0.5
Lcr
0.4
Scicode
0.3
Tau2
0.2
Hle
0.1
Terminalbench Hard
0.1

LLM Statsカテゴリスコア

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

価格設定

入力価格$0.2 / 1M tokens
出力価格$0.75 / 1M tokens
混合価格(3:1)$0.338 / 1M tokens

速度

トークン/秒126.2 tokens/s
初トークン遅延1.01s
初回答遅延16.86s

利用可能なプロバイダー

(LS内部単位)
プロバイダー入力価格出力価格
Novita200K1.0M
DeepInfra290K990K

外部リンク