DGEval - IMDG Code MCQ: leaderboard
The closed-book multiple-choice half of a dangerous-goods shipping certification, asked exactly as human candidates get it. Certified practitioners score 83.8 percent, which is the number every model on the board is measured against — and most of them are below it.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2027. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 95.8 | #58 (Gemini 3.1 Pro (Preview)) |
| 2 | Gemini 3.1 Pro (Preview) (Low) | 94.4 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Gemini 3.1 Pro (Preview) (Medium) | 94 | #58 (Gemini 3.1 Pro (Preview)) |
| 4 | GPT-5.5 (High) | 90.7 | #31 (GPT-5.5) |
| 5 | GPT-5.5 (Low) | 90.1 | #31 (GPT-5.5) |
| 6 | GPT-5.5 (Medium) | 89.6 | #31 (GPT-5.5) |
| 7 | Gemini 3.1 Flash Lite (Medium) | 79.2 | #191 (Gemini 3.1 Flash Lite) |
| 8 | Gemini 3.1 Flash Lite (High) | 79 | #191 (Gemini 3.1 Flash Lite) |
| 9 | GPT-5.4 Mini (High) | 76.1 | #212 (GPT-5.4 Mini) |
| 10 | Claude Sonnet 4.6 (Max) | 75.5 | #90 (Claude Sonnet 4.6) |
| 11 | GPT-5.4 Mini (Medium) | 74 | #212 (GPT-5.4 Mini) |
| 12 | Claude Opus 4.7 (Max) | 72.1 | #49 (Claude Opus 4.7) |
| 13 | GPT-5.5 (Non-reasoning) | 71.9 | #31 (GPT-5.5) |
| 14 | GPT-5.4 Mini (Low) | 71 | #212 (GPT-5.4 Mini) |
| 15 | Claude Opus 4.7 (Low) | 68.3 | #49 (Claude Opus 4.7) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=dgeval-imdg-code-mcq · How It Works · Data refreshed daily, snapshot 2026-10-20.