CapRiCorn-1K: leaderboard
Detailed audiovisual captioning of 1,000 manually annotated videos of 15 seconds to 10 minutes; CapRiCorn-1K (Chen et al., 2026). Keypoint accuracy is the share of annotated audio and visual keypoints that a GPT-4.1 judge finds mentioned in the caption, in percent; higher is better.
Metric: Keypoint accuracy (%; share of annotated keypoints a GPT-4.1 judge marks correctly mentioned in the audiovisual caption, 1,000 videos of 15 s to 10 min). Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 42.5 | #91 |
| 2 | Gemini 3 Flash | 41.5 | #159 |
| 3 | Qwen3 Omni 30B A3B Instruct | 10.3 | #902 |
| 4 | Qwen2.5-Omni-7B | 5.1 | #1797 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=capricorn-1k · How It Works · Data refreshed daily, snapshot 2026-11-12.