Qwen3 VL 32B Thinking

Name: Qwen3 VL 32B Thinking
Author: Alibaba

Multimodal

Alibaba

Qwen3-VL is a large multimodal model combining vision, language, and reasoning to achieve human-level perception and cognitive abilities. The 33B parameter Thinking version leads in multimodal reasoning and STEM tasks with OCR support, video understanding, and agentic interaction.

Key Specifications

Parameters

33.0B

Context

Release Date

September 21, 2025

Average Score

74.6%

API Documentation Repository Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

September 21, 2025

Last Update

February 17, 2026

Today

September 10, 2026

Technical Specifications

Parameters

33.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

General Knowledge

Tests on general knowledge and understanding

MMLU

Self-reported by the model provider • Self-reported

88.7%

Reasoning

Logical reasoning and analysis

GPQA

Self-reported by the model provider • Self-reported

73.1%

Multimodal

Working with images and visual data

AI2D

TEST • Self-reported

88.9%

Other Tests

Specialized benchmarks

OCRBench

• Self-reported

85.5%

MM-MT-Bench

• Self-reported

83.0%

DocVQAtest

• Self-reported

96.0%

ScreenSpot

• Self-reported

96.0%

MMLU-Redux

• Self-reported

92.0%

MMBench-V1.1

EN_V1.1 • Self-reported

91.0%

AIME 2025

Self-reported by the model provider • Self-reported

83.7%

AndroidWorld_SR

Self-reported by the model provider • Self-reported

63.7%

Arena-Hard v2

Self-reported by the model provider • Self-reported

60.5%

BFCL-v3

Self-reported by the model provider • Self-reported

71.7%

BLINK

Self-reported by the model provider • Self-reported

68.5%

CharadesSTA

Self-reported by the model provider • Self-reported

62.8%

CharXiv-D

Self-reported by the model provider • Self-reported

90.2%

CharXiv-R

Self-reported by the model provider • Self-reported

65.2%

Creative Writing v3

Self-reported by the model provider • Self-reported

83.3%

ERQA

Self-reported by the model provider • Self-reported

52.3%

Hallusion Bench

Self-reported by the model provider • Self-reported

67.4%

IFEval

Self-reported by the model provider • Self-reported

87.8%

Include

Self-reported by the model provider • Self-reported

76.3%

InfoVQAtest

Self-reported by the model provider • Self-reported

89.2%

LiveBench 20241125

Self-reported by the model provider • Self-reported

74.7%

LiveCodeBench v6

Self-reported by the model provider • Self-reported

65.6%

LVBench

Self-reported by the model provider • Self-reported

62.6%

MathVision

Self-reported by the model provider • Self-reported

70.2%

MathVista-Mini

testmini • Self-reported

85.9%

MMLU-Pro

Self-reported by the model provider • Self-reported

82.1%

MMLU-ProX

Self-reported by the model provider • Self-reported

77.2%

MMMU-Pro

full • Self-reported

68.1%

MMMU (val)

Self-reported by the model provider • Self-reported

78.1%

MMStar

Self-reported by the model provider • Self-reported

79.4%

MuirBench

Self-reported by the model provider • Self-reported

80.3%

Multi-IF

Self-reported by the model provider • Self-reported

78.0%

MVBench

Self-reported by the model provider • Self-reported

73.2%

OCRBench-V2 (en)

Self-reported by the model provider • Self-reported

68.4%

OCRBench-V2 (zh)

Self-reported by the model provider • Self-reported

62.1%

OSWorld

Self-reported by the model provider • Self-reported

41.0%

PolyMATH

Self-reported by the model provider • Self-reported

52.0%

RealWorldQA

Self-reported by the model provider • Self-reported

78.4%

ScreenSpot Pro

Self-reported by the model provider • Self-reported

57.1%

SimpleQA

Self-reported by the model provider • Self-reported

55.4%

SuperGPQA

Self-reported by the model provider • Self-reported

59.0%

VideoMME w/o sub.

Self-reported by the model provider • Self-reported

77.3%

VideoMMMU

Self-reported by the model provider • Self-reported

79.0%

WritingBench

Self-reported by the model provider • Self-reported

86.2%

License & Metadata

License

apache-2.0

Announcement Date

September 21, 2025

Last Updated

February 17, 2026

Similar Models

All Models

Qwen3.6-27B

Alibaba

MM27.8B

Best score:0.9 (GPQA)

Released:Apr 2026

Price:$0.60/1M tokens

Qwen3.6-35B-A3B

Alibaba

MM35.0B

Best score:0.9 (GPQA)

Released:Apr 2026

Qwen3.8-27B

Alibaba

MM27.8B

Best score:0.9 (GPQA)

Released:Aug 2026

Qwen2-VL-72B-Instruct

Alibaba

MM73.4B

Released:Aug 2024

Qwen2.5 VL 72B Instruct

Alibaba

MM72.0B

Released:Jan 2025

QvQ-72B-Preview

Alibaba

MM73.4B

Released:Dec 2024

Qwen3.5-4B

Alibaba

MM4.0B

Best score:0.8 (GPQA)

Released:Mar 2026

Qwen2.5 VL 32B Instruct

Alibaba

MM33.5B

Best score:0.9 (HumanEval)

Released:Feb 2025

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.