ChartBench: leaderboard
ChartBench evaluates complex visual reasoning over charts across chart types and task combinations.
Metric: Five-task average, Acc+ and NumberQA (%). Source: chartbench.github.io. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 59.45 | #343 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=chartbench · How It Works · Data refreshed daily, snapshot 2026-10-20.