GPTNT - Manual VQA Procedural Reasoning: leaderboard
Metric: Accuracy (%; 499 questions applying the manual's rules to a described bomb state to choose the Defuser's action). Source: arxiv.org. Saturation forecast: Around December 2026. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 (Non-reasoning) | 67.1 |
| 2 | Gemini 3 Flash (Preview) (Minimal) | 64.1 |
| 3 | GPT-5.2 (Non-reasoning) | 61.5 |
| 4 | Qwen 3.5 27B (Non-reasoning) | 53.7 |
Interactive version: theaggregate.ai/benchmark?slug=gptnt-manual-vqa-procedural-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-26.