Chartography (with tools)
Multimodal · 2026-07-16
Chartography's 100 domain-native chart-interpretation tasks answered with tools available, such as a Python sandbox with standard libraries or an image-cropping tool. Developers choose and often do not itemize the tools, and tool runs score several times higher than the no-tools runs on Surge AI's leaderboard, so they are kept apart; runs without tools are scored under Chartography.
Benchmark source: Surgehq
Top models (higher is better)
| Model | Score | Source |
|---|---|---|
| Kimi K3.1 Pro | 90.2 | Kimi K3.1 Pro launch post · Moonshot AI |
| Claude Opus 5.5 | 89.0 | Claude Opus 5.5 launch page · Anthropic |
| Claude Fable 5 | 85.2 | Claude Opus 5 system card · Anthropic |
| Claude Opus 5 | 84.0 | DeepSeek-V4.1-Flash model card · DeepSeek |
| DeepSeek V4.1 Flash 1017 | 83.0 | Kimi K3.1 Pro launch post · Moonshot AI |
| GPT-5.6 Sol | 79.9 | DeepSeek-V4.1-Flash model card · DeepSeek |
| DeepSeek V4.1 Flash | 78.9 | DeepSeek-V4.1-Flash model card · DeepSeek |
| Opus 4.8 | 75.0 | Claude Opus 5 system card · Anthropic |
| Kimi K3 | 68.1 | DeepSeek-V4.1-Flash model card · DeepSeek |
| DeepSeek V4 Flash Vision Exp | 64.3 | DeepSeek-V4-Flash-Vision-Exp model card · DeepSeek |