LLM Stats (LongVideoBench): leaderboard
LLM Stats mirror of LongVideoBench, evaluating question answering over long videos and their interleaved subtitles.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 2.1 Pro | 80.6 | #73 |
| 2 | Seed 2.1 Turbo | 80.6 | #93 |
| 3 | Kimi K2.5 | 79.8 | #152 |
| 4 | Seed 1.8 | 77.4 | #143 |
| 5 | Seed 2.0 Mini | 74.8 | #279 |
| 6 | Qwen 2.5 VL 7B Instruct | 54.7 | #650 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-longvideobench · How It Works · Data refreshed daily, snapshot 2026-11-05.