Which is better, GPT-4o or Qwen2.5 VL 32B Instruct?

Refresh Models —Deals —Regulatory —Sources —

Questions this page answers

Which is better, GPT-4o or Qwen2.5 VL 32B Instruct?: Across 14 shared benchmarks, GPT-4o scores higher on 7 and Qwen2.5 VL 32B Instruct on 7. The widest gap is GPQA Diamond, where GPT-4o scores 70.1 against 46.

GPT-4o vs Qwen2.5 VL 32B Instruct

Across 14 shared benchmarks, GPT-4o scores higher on 7 and Qwen2.5 VL 32B Instruct on 7. The widest gap is GPQA Diamond, where GPT-4o scores 70.1 against 46.

OpenAIvsAlibaba14 shared benchmarks7–7 head-to-head

Benchmark	GPT-4o	Qwen2.5 VL 32B Instruct
arena_vision	1162	1153
docvqa	92.8	94.8
GPQA Diamond	70.1	46
humaneval	90.6	91.5
MATH	85.3	82.2
mathvision	30.4	40
MathVista	63.8	74.7
mmlu	88.1	78.4
MMLU-Pro	74.7	68.8
MMMU	72.2	70
MMMU-Pro	59.9	49.5
MMStar	63.9	69.5
OSWorld-Verified	5	5.9
Video-MME	77.2	77.9

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.