Benchmarks / Reasoning

Chartography

Can a model read the charts professionals rely on — Kaplan-Meier curves, candlesticks, contour maps, Sankey diagrams, Bode plots, wind roses — from the image alone, to an expert-set precision? Maintained by Surge AI. Score: The board's accuracy: the share of answers inside the expert-set acceptable range of the golden answer, judged per answer by Gemini 3.5 Flash, averaged over the 100 tasks and ten runs per task.

41models
–Open Index weight
–last updated

Chartography

Higher is better · best effort per model

Chart settings
41 of 41 selected

full table below

#ModelEffortScoreCost / taskHarness
1GPT-6 Astramax71.0%––
2Opus 5.5max66.3%––
3GPT-6 Solmax53.6%––
4Fable 5.1max46.2%––
5GPT-5.6 Solmax45.0%––
6Gemini 3.7 Flashmedium43.0%––
7Gemini 3.8 Flashmedium42.5%––
8Gemini 3.5 Flashmedium35.9%––
9Fable 5max34.8%––
10Gemini 3.6 Flashmedium34.0%––
11GPT-5.6 Terramax34.0%––
12Muse Spark 1.2xhigh32.1%––
13GPT-5.6 Lunamax31.4%––
14GPT-5.5xhigh31.0%––
15GPT-5.4xhigh29.6%––
16GPT-6 Lunamax29.1%––
17Qwen 3.8 Maxxhigh29.1%––
18Muse Spark 1.3xhigh27.6%––
19Opus 5max27.3%––
20Kimi K3max26.6%––
21Gemini 3.1 Prohigh26.1%––
22Muse Spark 1.1xhigh24.4%––
23Grok 4.6high21.9%––
24Grok 4.7high17.4%––
25Qwen 3.8 Flashxhigh17.4%––
26Grok 4.5high17.0%––
27Sonnet 5max16.6%––
28Opus 4.7max16.5%––
29GLM 5.3 Flashmax16.3%––
30Opus 4.8max15.9%––
31Qwen 3.5 Plusnone15.9%––
32Qwen 3.7 Plushigh15.8%––
33Grok 4.3high12.9%––
34Kimi K2.5max12.6%––
35Kimi K2.6max12.2%––
36DeepSeek V4.1 Flashmax11.6%––
37DeepSeek V4 Flashmax11.2%––
38Gemini 3.5 Flash-Liteminimal10.2%––
39Inklinghigh10.0%––
40Inkling Smallxhigh9.1%––
41Mistral Large 3none9.0%––