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

#ModelScore
1Qwen 3.6 35B A3B53.7
2GPT-5.552.12
3Qwen 3.5 9B50.73
4Qwen 3.5 35B A3B50.35
5Qwen 3.7 Plus48.89
6Qwen 3.5 4B48.07
7Qwen 3.6 Plus45.79
8Gemma 4 26B A4B44.65
9Qwen 3 VL 32B44.21
10MiniCPM-V-2.635.53
11Ministral 3 14B30.4
12Gemma 4 E4B28.82

Interactive version: theaggregate.ai/benchmark?slug=mmhbench-first-person · How It Works · Data refreshed daily, snapshot 2026-09-29.