ChartAct - Box Charts: leaderboard

ChartAct (2026) questions on interactive box charts, pooled over the clean-page and dashboard environments: a multimodal agent acting on screenshots must reveal the evidence through chart interactions. Success rate in percent; higher is better.

Metric: Success rate (%) on box chart questions, pooling both environments; the model acts in an OSWorld-style desktop on screenshots only (pyautogui actions, at most 15 steps, temperature 0.3) to hover, click, zoom or drag 673 real interactive web charts and answers 1,440 human-reviewed questions, an LLM judge with majority voting deciding correctness; Claude Opus 4.7, GPT-5.5 and Gemini 3.1 Pro are scored on a representative 300-case subset per environment; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Opus 4.786.2#85
2Kimi K2.5Non-reasoning72.8#266
3GPT-5.567.2#54
4Doubao-Seed-2-0-Pro-260215High53.3
5Gemini 3.1 Pro (Preview)Low46.6#99
6Qwen 3.6 PlusThinking45.7#238
7Qwen 3.5 122B A10B45.3#346
8Qwen 3.5 35B A3B34.1#492
9GLM-4.6V33.3#710
10Gemma 4 31B (IT)26.1#350
11Qwen 3 VL PlusThinking20.3#512

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=chartact-box-charts · How It Works · Data refreshed daily, snapshot 2026-11-21.