Passer au contenu principal

Qwen3 VL 8B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Usage Commercial

Description

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.

Date de sortie
2025-10-14
Paramètres
9.0B
Longueur du contexte
262K
Modalités
image, text

Radar de capacités

80
general
0
coding
70
reasoning
60
scienceest.
60
agents
100
multimodal

Science est estimé à partir des scores scientifiques de LLM Stats ou du raisonnement lorsque les benchmarks scientifiques dédiés ne sont pas disponibles.

Classements

Domaine#RangScoreSource
Classement multimodal127
34.0
LS

Scores de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

Chat

MM-MT-Bench8.00 / 100Aut.
IFEvalGoogle Research (2023)83.2%Aut.
Multi-IF75.1%Aut.

Creativity

Creative Writing v382.4%Aut.

Factuality

SimpleQA49.6%Aut.

General

MMLU85.2%Aut.
MLVU-M75.1%Aut.
Include69.5%Aut.
BFCL-v363.0%Aut.
Arena-Hard v251.1%Aut.

Language

MMLU-Redux88.8%Aut.
MMLU-Pro77.3%Aut.
MMLU-ProX70.7%Aut.

Math

MathVista-Mini81.4%Aut.
AIME 202580.3%Aut.
LiveBench 2024112569.8%Aut.
MathVision62.7%Aut.
HMMT2560.6%Aut.
PolyMATH47.5%Aut.

Multimodal

VideoMMMU72.8%Aut.
Video-MME71.8%Aut.
OSWorld33.9%Aut.

Reasoning

CharXiv-D85.9%Aut.
GPQANYU + Cohere + Anthropic (2023)69.9%Aut.
LiveCodeBench v658.6%Aut.
CharXiv-R53.0%Aut.
SuperGPQA51.2%Aut.

Video

CharadesSTA59.9%Aut.

Vision

DocVQAtest95.3%Aut.
ScreenSpot93.6%Aut.
MMBench-V1.187.5%Aut.
InfoVQAtest86.0%Aut.
AI2D84.9%Aut.
OCRBench81.9%Aut.
MuirBench76.8%Aut.
CC-OCR76.3%Aut.
MMStar75.3%Aut.
MMMU (val)74.1%Aut.
RealWorldQA73.5%Aut.
MVBench69.0%Aut.
BLINK68.7%Aut.
Hallusion Bench65.4%Aut.
OCRBench-V2 (en)63.9%Aut.
MMMU-Pro60.4%Aut.
OCRBench-V2 (zh)59.2%Aut.
LVBench55.8%Aut.
ERQA46.8%Aut.
ScreenSpot Pro46.6%Aut.
ODinW39.8%Aut.

Writing

WritingBench85.5%Aut.

Indices d'évaluation AA

(Artificial Analysis)

Aucune donnée d'évaluation AA disponible

Scores par catégorie LLM Stats

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

Tarification

Prix d'entrée$0.25 / 1M tokens
Prix de sortie$0.75 / 1M tokens
Prix mixte (3:1)$0.375 / 1M tokens
Prix de lecture cache$0.12 / 1M tokens

Vitesse

Aucune donnée de vitesse disponible

Classement des Prix par Fournisseur

Classement des Prix par Fournisseur

3 fournisseurs

Moins cher: AlibabaPlus cher: Kilo Gateway
FournisseurEntréeSortie
1AlibabaPRINCIPAL
$0.25
$0.75
2OpenRouter
$0.25
$0.75
3Kilo Gateway
$0.25
$0.75

Comparer les prix entre différents fournisseurs API pour ce modèle.

Sources externes