TRAP - Multimodal Task Accuracy: leaderboard
Metric: Task accuracy (%) on the 200 multimodal-document scenarios of 500 document-grounded scenarios (100 text, 200 image, 200 multimodal) over ten private document types; each pairs a task query that needs a tool call with private fields and an attack query eliciting the same fields in natural language; rule-based evaluator, greedy decoding. Source: arxiv.org. Saturation forecast: Around 2029. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini | 73 |
| 2 | GPT-5 | 72 |
| 3 | Qwen 3 VL 32B | 71 |
| 4 | Gemini 2.5 Flash | 69 |
| 5 | GPT-5.4 Mini | 67.5 |
| 6 | Gemini 2.5 Flash Lite | 66.5 |
| 7 | InternVL3.5-8B | 64 |
| 8 | GPT-4o Mini | 60 |
| 9 | Claude Sonnet 4.5 | 55 |
| 10 | Claude Haiku 4.5 | 55 |
| 11 | Gemini 2.5 Pro | 52 |
Interactive version: theaggregate.ai/benchmark?slug=trap-multimodal-task-accuracy · How It Works · Data refreshed daily, snapshot 2026-09-29.