DVBench - Chart Reasoning: leaderboard
DVBench accuracy on insights that must be inferred across chart elements and frames of a data video, such as trends, comparisons and counts.
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) | 77.87 | #64 |
| 2 | Kimi K2.5 | 75.41 | #152 |
| 3 | Claude Sonnet 4.6 | 71.58 | #94 |
| 4 | GPT-5.4 | 62.57 | #83 |
| 5 | Qwen 3.5 27B | 56.83 | #192 |
| 6 | Gemma 4 31B (IT) | 56.01 | #178 |
| 7 | Qwen 3.5 397B A17B | 54.92 | #154 |
| 8 | Qwen 3.5 122B A10B | 52.73 | #184 |
| 9 | Qwen 3.5 9B | 47.27 | #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-chart-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.