NARU: leaderboard
NARU (2026) is a video question-answering benchmark on 155 Japanese long-form videos (146.8 hours) with 1,481 four-choice questions on narrative evolution and cultural nuance. Accuracy in percent over all questions; higher is better. Each model uses the recommended input of its family: Gemini at 0.25 frames per second through native video APIs, open models at the largest frame count their context allows.
Metric: Accuracy (%; all 1,481 questions; four-choice questions over 155 Japanese long-form videos (146.8 hours), each model with its family's recommended input (Gemini through native video APIs at 0.25 fps, open models at the largest uniform frame count their context allows)). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 76.2 | #159 |
| 2 | Gemini 3 Pro | 70 | #135 |
| 3 | Gemini 2.5 Flash | 51.4 | #452 |
| 4 | Qwen 3.5 9B | 39.8 | #803 |
| 5 | Qwen 2.5 VL 7B | 29.7 | #1796 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=naru · How It Works · Data refreshed daily, snapshot 2026-11-12.