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

34
general
65
coding
86
reasoning
57
science
70
agents
100
multimodal

Rankings

Dominio#PosiciónPuntuaciónFuente
Ranking de codificación281
53.0
AA
Ranking general235
50.0
AA
Ranking multimodal82
51.0
LS
Ciencia285
48.0
AA

Puntuaciones de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

Chat

MM-MT-Bench8.50 / 100Aut.
IFEvalGoogle Research (2023)88.2%Aut.
Multi-IF79.1%Aut.

Creativity

Creative Writing v385.7%Aut.

Factuality

SimpleQA44.4%Aut.

General

MMLU90.6%Aut.
Include80.0%Aut.
SIFO0.77 / 100Aut.
BFCL-v371.9%Aut.
SIFO-Multiturn0.71 / 100Aut.

Language

MMLU-Redux93.7%Aut.
MMLU-Pro83.8%Aut.
MMLU-ProX80.6%Aut.

Math

AIME 202589.7%Aut.
MathVista-Mini85.8%Aut.
MathVerse-Mini0.85 / 100Aut.
LiveBench 2024112579.6%Aut.
HMMT2577.4%Aut.
MathVision74.6%Aut.

Multimodal

VideoMMMU80.0%Aut.
VideoMME w/o sub.79.0%Aut.
OSWorld-G0.68 / 100Aut.
OSWorld38.1%Aut.

Reasoning

ZebraLogic97.3%Aut.
LiveCodeBench v670.1%Aut.
CharXiv-R66.1%Aut.
SuperGPQA64.3%Aut.
Humanity's Last Exam13.6%Aut.

Video

MLVU83.8%Aut.
CharadesSTA63.5%Aut.

Vision

DocVQAtest96.5%Aut.
ScreenSpot95.4%Aut.
CountBench0.94 / 100Aut.
Design2Code0.93 / 100Aut.
MIABench0.93 / 100Aut.
RefCOCO-avg0.92 / 100Aut.
MMBench-V1.190.6%Aut.
InfoVQAtest89.5%Aut.
AI2D89.2%Aut.
OCRBench87.5%Aut.
EmbSpatialBench0.84 / 100Aut.
CC-OCR81.5%Aut.
RealWorldQA81.3%Aut.
MMMUval80.6%Aut.
MuirBench80.1%Aut.
MMStar78.7%Aut.
RoboSpatialHome0.74 / 100Aut.
Objectron0.71 / 100Aut.
RefSpatialBench0.70 / 100Aut.
MMMU-Pro69.3%Aut.
BLINK67.1%Aut.
OCRBench-V2 (en)66.8%Aut.
Hallusion Bench66.7%Aut.
LVBench63.6%Aut.
OCRBench-V2 (zh)63.5%Aut.
ScreenSpot Pro61.8%Aut.
SimpleVQA0.61 / 100Aut.
MMLongBench-Doc0.56 / 100Aut.
ARKitScenes0.54 / 100Aut.
ERQA52.5%Aut.
ODinW43.2%Aut.
SUNRGBD0.35 / 100Aut.
VisuLogic0.34 / 100Aut.
ZEROBench-Sub0.28 / 100Aut.
Hypersim0.11 / 100Aut.
ZEROBench0.04 / 100Aut.

Writing

WritingBench86.7%Aut.

Índices de evaluación AA

(Artificial Analysis)
Aime 25(MAA (Mathematical Association of America))
88.3
Math Index(Artificial Analysis)
88.3
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
83.6
Gpqa(NYU + Cohere + Anthropic (2023))
77.2
Livecodebench(UC Berkeley + MIT + Cornell (2024))
64.6
Lcr(Artificial Analysis)
63.7
Ifbench(Google Research (2023))
56.5
Tau2(Sierra + U Toronto + Vector Institute (2025))
54.1
Intelligence Index(Artificial Analysis)
13.4
Hle(Center for AI Safety + Scale AI (2025))
11.9
Terminalbench Hard(Stanford × Laude Institute (2026))
11.4

Puntuaciones por categoría LLM Stats

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

Precios

Precio de entrada$0.4 / 1M tokens
Precio de salida$4 / 1M tokens
Precio mixto (3:1)$1.3 / 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

11 proveedores

Más barato: Venice AIMás caro: NovitaAI
ProveedorEntradaSalida
1Venice AIMás barato
$0.21
$1.9
2Alibaba (China)
$0.28671
$1.14682
3Merge Gateway
$0.287
$2.867
4AlibabaPRINCIPAL
$0.4
$4
5OpenRouter
$0.4
$4
6Kilo Gateway
$0.4
$4
7OrcaRouter
$0.4
$4
8Eden AI
$0.4
$4
9NanoGPT
$0.5
$6
10Cortecs
$0.617
$3.119
11NovitaAI
$0.98
$3.95

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas