Saltar al contenido principal

Qwen3 VL 235B A22B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Uso Comercial

Descripción

Qwen3-VL-235B-A22B-Thinking is the most powerful vision-language model in the Qwen series, featuring 236B parameters with MoE architecture for reasoning-enhanced multimodal understanding. Key capabilities include: Visual Agent (operates PC/mobile GUIs, recognizes elements, invokes tools), Visual Coding (generates Draw.io/HTML/CSS/JS from images/videos), Advanced Spatial Perception (2D grounding and 3D grounding for spatial reasoning and embodied AI), Long Context & Video Understanding (native 256K context expandable to 1M, handles hours-long video with second-level indexing), Enhanced Multimodal Reasoning (excels in STEM/Math with causal analysis), Upgraded Visual Recognition (celebrities, anime, products, landmarks, flora/fauna), and Expanded OCR (32 languages, robust in low light/blur/tilt). Architecture innovations include Interleaved-MRoPE for positional embeddings, DeepStack for multi-level ViT feature fusion, and Text-Timestamp Alignment for precise video temporal modeling.

Fecha de lanzamiento
2025-09-23
Parámetros
236.0B
Longitud del contexto
131K
Modalidades
image, text, video

Radar de capacidades

38
general
59
coding
86
reasoning
51
science
70
agents
100
multimodal

Rankings

Dominio#PosiciónPuntuaciónFuente
Capacidad agéntica54
52.0
LS
Ranking de codificación204
54.0
AA
Ranking general209
55.0
AA
Ranking multimodal33
55.0
LS
Ciencia206
55.0
AA

Puntuaciones de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

3d

Objectron0.71 / 100Aut.
BLINK67.1%Aut.
ARKitScenes0.54 / 100Aut.
SUNRGBD0.35 / 100Aut.
Hypersim0.11 / 100Aut.

Agents

SIFO0.77 / 100Aut.
BFCL-v371.9%Aut.
SIFO-Multiturn0.71 / 100Aut.
OSWorld-G0.68 / 100Aut.
OSWorld38.1%Aut.

Chemistry

SuperGPQA64.3%Aut.

Code

Design2Code0.93 / 100Aut.

Communication

MM-MT-Bench8.50 / 100Aut.
WritingBench86.7%Aut.
Multi-IF79.1%Aut.

Creativity

Creative Writing v385.7%Aut.

Embodied

EmbSpatialBench0.84 / 100Aut.
RoboSpatialHome0.74 / 100Aut.

Factuality

SimpleQA44.4%Aut.

Finance

MMLU90.6%Aut.
MMLU-Pro83.8%Aut.
MMLU-ProX80.6%Aut.

General

MMLU-Redux93.7%Aut.
IFEvalGoogle Research (2023)88.2%Aut.
MMMUval80.6%Aut.
Include80.0%Aut.
LiveBench 2024112579.6%Aut.
MMStar78.7%Aut.
LiveCodeBench v670.1%Aut.
MMMU-Pro69.3%Aut.
SimpleVQA0.61 / 100Aut.

Grounding

ScreenSpot95.4%Aut.
RefCOCO-avg0.92 / 100Aut.
RefSpatialBench0.70 / 100Aut.
ScreenSpot Pro61.8%Aut.

Healthcare

VideoMMMU80.0%Aut.

Image To Text

OCRBench87.5%Aut.
OCRBench-V2 (en)66.8%Aut.
OCRBench-V2 (zh)63.5%Aut.

Instruction Following

MIABench0.93 / 100Aut.

Language

CharadesSTA63.5%Aut.

Long Context

MLVU83.8%Aut.
LVBench63.6%Aut.
MMLongBench-Doc0.56 / 100Aut.

Math

AIME 202589.7%Aut.
MathVista-Mini85.8%Aut.
MathVerse-Mini0.85 / 100Aut.
HMMT2577.4%Aut.
MathVision74.6%Aut.
Humanity's Last Exam13.6%Aut.

Multimodal

DocVQAtest96.5%Aut.
MMBench-V1.190.6%Aut.
InfoVQAtest89.5%Aut.
AI2D89.2%Aut.
CC-OCR81.5%Aut.
MuirBench80.1%Aut.
VideoMME w/o sub.79.0%Aut.
CharXiv-R66.1%Aut.
VisuLogic0.34 / 100Aut.
ZEROBench-Sub0.28 / 100Aut.
ZEROBench0.04 / 100Aut.

Reasoning

ZebraLogic97.3%Aut.
CountBench0.94 / 100Aut.
Hallusion Bench66.7%Aut.
ERQA52.5%Aut.

Spatial Reasoning

RealWorldQA81.3%Aut.

Vision

ODinW43.2%Aut.

Índices de evaluación AA

(Artificial Analysis)
Math Index(Artificial Analysis)
88.3
Intelligence Index(Artificial Analysis)
20.9
Aime 25(MAA (Mathematical Association of America))
0.9
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
0.8
Gpqa(NYU + Cohere + Anthropic (2023))
0.8
Livecodebench(UC Berkeley + MIT + Cornell (2024))
0.6
Lcr(Artificial Analysis)
0.6
Ifbench(Google Research (2023))
0.6
Tau2(Sierra + U Toronto + Vector Institute (2025))
0.5
Scicode(UIUC + Argonne National Lab (2024))
0.4
Hle(Center for AI Safety + Scale AI (2025))
0.1
Terminalbench Hard(Stanford × Laude Institute (2026))
0.1

Puntuaciones por categoría LLM Stats

(LLM Stats (zeroeval))
Communication
3
Multimodal
100
Creativity
90
Writing
90
Legal
80
Math
80
Structured Output
80
Instruction Following
80
Language
80
Embodied
80
Finance
80
General
80
Grounding
80
Healthcare
80
Text-to-image
80
Video
80
Long Context
70
Reasoning
70
Spatial Reasoning
70
Image To Text
70
Tool Calling
70
Vision
70
Physics
60
Agents
60
Chemistry
60
Economics
60
Factuality
40
3d
40

Precios

Precio de entrada$0.7 / 1M tokens
Precio de salida$8.4 / 1M tokens
Precio mixto (3:1)$2.625 / 1M tokens

Velocidad

Tokens/seg0.0
Retraso del primer token0.00s
Tiempo hasta la respuesta0.00s

Ranking de Precios por Proveedor

Ranking de Precios por Proveedor

7 proveedores

Más barato: Venice AIMás caro: NovitaAI
ProveedorEntradaSalida
1Venice AIMás barato
$0.21
$1.9
2Alibaba (China)
$0.28671
$1.14682
3OpenRouter
$0.4
$4
4Kilo Gateway
$0.4
$4
5Cortecs
$0.617
$3.119
6AlibabaPRINCIPAL
$0.7
$8.4
7NovitaAI
$0.98
$3.95

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas