ChartMuseum — leaderboard
Chart-understanding benchmark reported in Anthropic system cards, with no-tools and Python-tools variants.
Metric: Overall Accuracy (%). Source: chartmuseum-leaderboard.github.io. Status: saturation imminent. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 80.7 |
| 2 | Gemini 3 Flash | 78.6 |
| 3 | GPT-5.2 (High) | 77 |
| 4 | Gemma 4 31B (IT) | 71.1 |
| 5 | Gemma 4 26B A4B (IT) | 70.6 |
| 6 | GPT-5 Mini (High) | 63.3 |
| 7 | Gemini 2.5 Pro | 63 |
| 8 | GPT-5 (High) | 62.9 |
| 9 | O4 Mini (High) | 61.5 |
| 10 | O3 (High) | 60.9 |
| 11 | Claude Opus 4.5 (High) | 60.7 |
| 12 | Claude Opus 4.1 | 59.1 |
| 13 | Claude Sonnet 4 | 52.6 |
| 14 | Qwen 3 VL 30B A3B (Thinking) | 49.7 |
| 15 | Qwen 3 VL 8B (Thinking) | 44.4 |
Interactive version: theaggregate.ai/benchmark?slug=chartmuseum · How the rankings work · Data refreshed daily, snapshot 2026-07-22.