ChartBench - NumberQA: leaderboard
ChartBench NumberQA (Xu et al., 2023): the model reads a numeric value off a chart, most without printed data labels, and an answer counts as correct within 5% of the true value (an LLM extracts the number from the response). Accuracy in percent; higher is better.
Metric: Accuracy within 5% of the value (%). Source: chartbench.github.io. Saturation forecast: Around February 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 40.48 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chartbench-numberqa · How It Works · Data refreshed daily, snapshot 2026-11-12.