MMHBench - First-Person: leaderboard
Metric: Accuracy (%; the 1,579 first-person perspective-taking multiple-choice questions on 268 long-form mental-health-related videos, 64 uniformly sampled frames, zero-shot, asked in the voice of a social role such as the person with lived experience, a mental health professional or a family member; higher is better). Source: arxiv.org. Saturation forecast: Around 2030. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.6 35B A3B | 53.7 |
| 2 | GPT-5.5 | 52.12 |
| 3 | Qwen 3.5 9B | 50.73 |
| 4 | Qwen 3.5 35B A3B | 50.35 |
| 5 | Qwen 3.7 Plus | 48.89 |
| 6 | Qwen 3.5 4B | 48.07 |
| 7 | Qwen 3.6 Plus | 45.79 |
| 8 | Gemma 4 26B A4B | 44.65 |
| 9 | Qwen 3 VL 32B | 44.21 |
| 10 | MiniCPM-V-2.6 | 35.53 |
| 11 | Ministral 3 14B | 30.4 |
| 12 | Gemma 4 E4B | 28.82 |
Interactive version: theaggregate.ai/benchmark?slug=mmhbench-first-person · How It Works · Data refreshed daily, snapshot 2026-09-29.