DVBench - Narrative: leaderboard
DVBench accuracy on questions about a data video's storytelling logic, topic, intent and communicative stance.
Metric: Accuracy (%). Source: bomiaowang.github.io. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 76.57 | #64 |
| 2 | Claude Sonnet 4.6 | 75.43 | #94 |
| 3 | Qwen 3.5 27B | 74.29 | #192 |
| 4 | Qwen 3.5 397B A17B | 73.14 | #154 |
| 5 | Kimi K2.5 | 72.57 | #152 |
| 6 | Qwen 3.5 122B A10B | 70.86 | #184 |
| 7 | GPT-5.4 | 69.14 | #83 |
| 8 | Gemma 4 31B (IT) | 64 | #178 |
| 9 | Qwen 3.5 9B | 55.43 | #391 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=dvbench-narrative · How It Works · Data refreshed daily, snapshot 2026-11-05.