VRIQ (Abstract): leaderboard
Metric: Mean accuracy (%) over the five abstract-domain VRIQ categories (1,000 four-option IQ-style puzzles over geometric primitives: sequence completion, 3D visualization, figure rotation, odd one out, matrix prediction; 200 each); higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 35.5 | #287 (Qwen 3 VL 32B) |
| 2 | Qwen 2.5 VL 7B Instruct | 31.2 | #643 |
| 3 | GPT-4o | 30.3 | #333 |
| 4 | GPT-5.1 | 29.3 | #131 |
| 5 | Gemini 2.5 Pro | 28.7 | #145 |
| 6 | GPT-4o Mini | 25.1 | #588 |
| 7 | GPT-5 Mini | 21.6 | #176 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=vriq-abstract · How It Works · Data refreshed daily, snapshot 2026-10-11.