The AI arena is free today

Open Superagent

Qwen2.5 VL 72B Instruct vs Qwen3 VL 32B Thinking

Qwen3 VL 32B Thinking leads the LLM Stats Score 23.6 to 12.7.

Alibaba Cloud / Qwen Team · Alibaba Cloud / Qwen Team · Updated for 2026

Which is better?

Qwen3 VL 32B Thinking leads the overall LLM Stats Score 23.6 to 12.7, ranking #173 overall.

In the 15 individual benchmarks reported for both models, Qwen3 VL 32B Thinking wins 14; this is a narrower head-to-head signal than the composite indexes.

Based on current LLM Stats indexes, shared benchmarks, pricing, and model metadata for 2026.

Choose Qwen2.5 VL 72B Instruct

  • you are already invested in the Alibaba Cloud / Qwen Team ecosystem

Choose Qwen3 VL 32B Thinking

  • overall performance matters — it scores 23.6 and ranks #173 on LLM Stats
  • your work emphasizes reasoning — it leads those capability indexes
  • you value its reported benchmark strengths — it wins 14 of 15 exact shared results
  • you want the most recent training data — it shipped Sep 2025

At a glance

The differences that matter most.

Qwen2.5 VL 72B Instruct
Alibaba Cloud / Qwen Team
Qwen3 VL 32B Thinking
Alibaba Cloud / Qwen Team
Core performance indexes
LLM Stats Score
12.7
#249
23.6
#173
Reasoning
12.8
#240
24.7
#153
Agents
4.7
#152
13.0
#96
Cost, coverage & limits
Benchmark wins
1 of 15
14 of 15
Input price
— / M
— / M
Output price
— / M
— / M
Context window
—
—

Capability indexes

Additional strengths measured across groups of related public benchmarks

3 shared
Index
Qwen2.5 VL 72B Instruct
Qwen3 VL 32B Thinking
Math
7.6#263
25.4#110
Vision
9.2#130
16.9#85
Multimodal
12.1#106
19.5#72
Conservative TrueSkill rating · higher is betterHow scores work

Individual benchmarks

30 reported for Qwen2.5 VL 72B Instruct · 47 for Qwen3 VL 32B Thinking

15 shared

Qwen2.5 VL 72B Instruct outperforms in 1 benchmarks (OCRBench), while Qwen3 VL 32B Thinking is better at 14 benchmarks (AI2D, AndroidWorld_SR, Hallusion Bench, LVBench, MathVision, MathVista-Mini, MMMU-Pro, MMStar, MVBench, OCRBench-V2 (en), OSWorld, ScreenSpot, ScreenSpot Pro, VideoMME w/o sub.).

Qwen3 VL 32B Thinking significantly outperforms across most benchmarks.

Fri Sep 11 2026 • llm-stats.com

Human preference

Blind head-to-head votes and playground preference scores

Model Size

Parameter count comparison

39.0B diff

Qwen2.5 VL 72B Instruct has 39.0B more parameters than Qwen3 VL 32B Thinking, making it 118.2% larger.

Alibaba Cloud / Qwen Team
Qwen2.5 VL 72B Instruct
72.0Bparameters
Alibaba Cloud / Qwen Team
Qwen3 VL 32B Thinking
33.0Bparameters
72.0B
Qwen2.5 VL 72B Instruct
33.0B
Qwen3 VL 32B Thinking

Input capabilities

Documented input modalities across available providers

Both Qwen2.5 VL 72B Instruct and Qwen3 VL 32B Thinking support multimodal inputs.

They are both capable of processing various types of data, offering versatility in application.

Qwen2.5 VL 72B Instruct

Text
Images
Audio
Video

Qwen3 VL 32B Thinking

Text
Images
Audio
Video

License

Usage and distribution terms

Qwen2.5 VL 72B Instruct is licensed under tongyi-qianwen, while Qwen3 VL 32B Thinking uses Apache 2.0.

License differences may affect how you can use these models in commercial or open-source projects.

Qwen2.5 VL 72B Instruct

tongyi-qianwen

Open weights

Qwen3 VL 32B Thinking

Apache 2.0

Open weights

Release Timeline

When each model was launched

Qwen2.5 VL 72B Instruct was released on 2025-01-26, while Qwen3 VL 32B Thinking was released on 2025-09-22.

Qwen3 VL 32B Thinking is 8 months newer than Qwen2.5 VL 72B Instruct.

Qwen2.5 VL 72B Instruct

Jan 26, 2025

1.6 years ago

Qwen3 VL 32B Thinking

Sep 22, 2025

11 months ago

7mo newer

Knowledge Cutoff

When training data ends

Neither model specifies a knowledge cutoff date.

Unable to compare the recency of their training data.

No cutoff dates available

Outputs Comparison

Notice missing or incorrect data?Start an Issue discussion→

Judge for yourself.

Run your own prompts against Qwen2.5 VL 72B Instruct and Qwen3 VL 32B Thinking side-by-side, then vote on the output you prefer.

Qwen2.5 VL 72B Instruct
✓ Preferred
Qwen3 VL 32B Thinking
Open in Playground

FAQ

Common questions about Qwen2.5 VL 72B Instruct vs Qwen3 VL 32B Thinking.

Which is better, Qwen2.5 VL 72B Instruct or Qwen3 VL 32B Thinking?

Qwen3 VL 32B Thinking leads the LLM Stats Score 23.6 to 12.7. Qwen2.5 VL 72B Instruct is made by Alibaba Cloud / Qwen Team and Qwen3 VL 32B Thinking is made by Alibaba Cloud / Qwen Team. The best choice depends on your use case — compare their capability indexes, individual benchmarks, pricing, and limits above.

How does Qwen2.5 VL 72B Instruct compare to Qwen3 VL 32B Thinking in benchmarks?

Qwen2.5 VL 72B Instruct scores DocVQA: 96.4%, Android Control Low_EM: 93.7%, ChartQA: 89.5%, OCRBench: 88.5%, AI2D: 88.4%. Qwen3 VL 32B Thinking scores DocVQAtest: 96.1%, ScreenSpot: 95.7%, MMLU-Redux: 91.9%, MMBench-V1.1: 90.8%, CharXiv-D: 90.2%.

What are the main differences between Qwen2.5 VL 72B Instruct and Qwen3 VL 32B Thinking?

Key differences include LLM Stats Score (12.7 vs 23.6), licensing (tongyi-qianwen vs Apache 2.0). See the full comparison above for benchmark-by-benchmark results.

Related comparisons