VNU-Bench - Multimodal Presentation Comparison: leaderboard

Metric: Accuracy (%; question type T5, 4-option multiple choice). Source: arxiv.org. Saturation forecast: Around August 2028. 13 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro66.54
2Qwen 3 VL 8B61.65
3Gemini 2.5 Flash61.22
4Claude Sonnet 4.560.57
5InternVL3-8B55.48
6Qwen 2.5 VL 7B Instruct55.2

Interactive version: theaggregate.ai/benchmark?slug=vnu-bench-multimodal-presentation-comparison · How It Works · Data refreshed daily, snapshot 2026-09-25.