HarmVideoBench - Beyond-Clip Reasoning: leaderboard
HarmVideoBench (2026) asks multimodal models four-option questions about 1,379 real harmful videos at three levels of understanding; API models watch the video with audio and open models see 12 key frames with the transcript. This board is accuracy in percent on questions that need context beyond the clip; chance is 25 and higher is better.
Metric: Beyond-Clip Reasoning accuracy (%) on four-option multiple-choice questions about 1,379 real harmful videos (one question per video per level), temperature 0 greedy decoding; API models take the video and audio natively, open models 12 sampled key frames plus the transcript; chance 25; questions that need context beyond the clip; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 65.8 | #94 |
| 2 | GPT-5.5 | 63.9 | #49 |
| 3 | Claude Opus 4.7 | 63.2 | #79 |
| 4 | Claude Sonnet 4.6 | 62.7 | #161 |
| 5 | Qwen 3 VL 235B A22B (Thinking) | 62.5 | #451 |
| 6 | Qwen 3 VL 32B (Thinking) | 62.1 | #559 |
| 7 | Qwen 3 VL 235B A22B Instruct | 62 | #522 |
| 8 | Qwen 3 VL 30B A3B (Thinking) | 61.2 | #712 |
| 9 | Qwen 3 VL 32B Instruct | 61.1 | #585 |
| 10 | Qwen 3 VL 8B (Thinking) | 59.5 | #887 |
| 11 | Qwen 3 VL 30B A3B Instruct | 58.5 | #871 |
| 12 | MiniCPM-V-4.6 | 53.3 | #3525 |
| 13 | LFM2.5-VL-1.6B | 51.4 | #3872 |
| 14 | GLM-4.5V (Non-reasoning) | 50.9 | #744 (GLM-4.5V) |
| 15 | Qwen 3 VL 8B Instruct | 48.1 | #981 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=harmvideobench-beyond-clip-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-13.