GroupToM-Bench - Intention: leaderboard
Metric: strict exact-match accuracy (%) over multi-select options (random 6.7) on Level 3 intention prediction (the behavioural strategy an agent will adopt), multiple choice; 240 multimodal multi-agent scenarios (scene image plus dialogue); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 73.9 |
| 2 | GPT-4o | 72.7 |
| 3 | GPT-5 | 72.3 |
| 4 | Claude Haiku 4.5 | 70 |
| 5 | GPT-5 Nano | 69.2 |
| 6 | GPT-5 Mini | 69.2 |
| 7 | InternVL3.5-8B | 64.2 |
| 8 | Qwen 2.5 VL 7B Instruct | 63.5 |
| 9 | Qwen 2 VL 7B | 50 |
Interactive version: theaggregate.ai/benchmark?slug=grouptom-bench-intention · How It Works · Data refreshed daily, snapshot 2026-09-29.