MMHBench - Third-Person: leaderboard
Metric: Accuracy (%; the 605 third-person assessment multiple-choice questions on 268 long-form mental-health-related videos, 64 uniformly sampled frames, zero-shot about observable behavior and multimodal evidence; higher is better). Source: arxiv.org. Saturation forecast: Around February 2028. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 63.64 |
| 2 | Qwen 3.7 Plus | 58.35 |
| 3 | Qwen 3.6 Plus | 58.18 |
| 4 | Qwen 3.6 35B A3B | 57.85 |
| 5 | Qwen 3.5 35B A3B | 55.37 |
| 6 | Qwen 3.5 9B | 53.22 |
| 7 | Qwen 3.5 4B | 51.9 |
| 8 | Qwen 3 VL 32B | 50.74 |
| 9 | Gemma 4 26B A4B | 45.95 |
| 10 | MiniCPM-V-2.6 | 43.31 |
| 11 | Ministral 3 14B | 30.08 |
| 12 | Gemma 4 E4B | 28.6 |
Interactive version: theaggregate.ai/benchmark?slug=mmhbench-third-person · How It Works · Data refreshed daily, snapshot 2026-09-29.