NARU - Aizuchi (Conversational Mechanics): leaderboard
Metric: Accuracy (%; the 143 cultural questions of category C.1; 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 December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 71.3 |
| 2 | Gemini 3 Pro | 60.1 |
| 3 | Qwen 3.5 9B | 41.3 |
| 4 | Gemini 2.5 Flash | 37.8 |
| 5 | Qwen 3 VL 8B | 32.2 |
| 6 | Qwen 2.5 VL 7B | 22.4 |
Interactive version: theaggregate.ai/benchmark?slug=naru-aizuchi-conversational-mechanics · How It Works · Data refreshed daily, snapshot 2026-09-26.