LongVideoBench: leaderboard
Long-form video understanding benchmark requiring models to reason over extended temporal contexts.
Metric: Test Total Accuracy (%). Source: longvideobench.github.io. Status: saturated. 35 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro | 64.4 |
| 2 | Gemini 1.5 Flash | 62.4 |
| 3 | GPT-4 Turbo | 60.7 |
| 4 | GPT-4o Mini | 58.8 |
| 5 | Qwen 2 VL 7B | 56.8 |
Interactive version: theaggregate.ai/benchmark?slug=longvideobench · How It Works · Data refreshed daily, snapshot 2026-09-05.