Перейти к основному содержанию

Qwen3 VL 235B A22B (Reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

Описание

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.

Дата выхода
2025-09-23
Параметры
236.0B
Длина контекста
131K
Модальности
image, text, video

Радар способностей

42
general
38
coding
86
reasoning
51
scienceоцен.
70
agents
100
multimodal

Science использует прокси на основе рассуждений, когда специализированные научные бенчмарки недоступны.

Рейтинги

Домен#МестоОценкаИсточник
Agents & Tools24
66.0
LS
Code Ranking171
47.0
AA
General Ranking146
59.0
AA
Math Reasoning49
89.0
AA
Multimodal Ranking64
67.0
LS
Reasoning37
75.0
LS
Science137
56.0
AA

Оценки бенчмарков (LLM Stats)

3d

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

Agents

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

Chemistry

SuperGPQA64.3%Сам.

Code

Design2Code0.93 / 100Сам.

Communication

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

Creativity

Creative Writing v385.7%Сам.

Embodied

EmbSpatialBench0.84 / 100Сам.
RoboSpatialHome0.74 / 100Сам.

Factuality

SimpleQA44.4%Сам.

Finance

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

General

MMLU-Redux93.7%Сам.
IFEval88.2%Сам.
MMMUval80.6%Сам.
Include80.0%Сам.
LiveBench 2024112579.6%Сам.
MMStar78.7%Сам.
LiveCodeBench v670.1%Сам.
MMMU-Pro69.3%Сам.
SimpleVQA0.61 / 100Сам.

Grounding

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

Healthcare

VideoMMMU80.0%Сам.

Image To Text

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

Instruction Following

MIABench0.93 / 100Сам.

Language

CharadesSTA63.5%Сам.

Long Context

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

Math

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

Multimodal

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

Reasoning

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

Spatial Reasoning

RealWorldQA81.3%Сам.

Vision

ODinW43.2%Сам.

Индексы оценки AA

Math Index
88.3
Intelligence Index
27.6
Coding Index
20.9
Aime 25
0.9
Mmlu Pro
0.8
Gpqa
0.8
Livecodebench
0.6
Lcr
0.6
Ifbench
0.6
Tau2
0.5
Scicode
0.4
Terminalbench Hard
0.1
Hle
0.1

Оценки категорий LLM Stats

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

Цены

Цена ввода$0.84 / 1M tokens
Цена вывода$6.175 / 1M tokens
Смешанная цена (3:1)$2.174 / 1M tokens

Скорость

Токенов/сек30.0 tokens/s
Задержка первого токена1.35s
Время до первого ответа68.10s

Доступные провайдеры

(Внутренние единицы LS)
ПровайдерЦена вводаЦена вывода
DeepInfra450K3.5M
Novita980K4.0M

Внешние ссылки