ChartArena - Pie Chart (Chinese): leaderboard

Metric: mAP-high (%) on Chinese pie charts, each model parses the chart image into its native structured format, normalized format-agnostically to (header, entity, value) triples for numeric charts or to labeled graphs and trees for flowcharts and mind maps; mean average precision over similarity thresholds at the high tolerance level, averaged over three visual styles (digital renderings, printed photos, hand-drawn photos); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 43 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)95.1
2Seed 2.0 Pro (Non-reasoning)90.5
3Qwen 3.5 35B A3B (Thinking)88.4
4Gemini 2.5 Pro87.3
5Gemma 4 31B (IT)87.2
6Kimi K2.5 (Non-reasoning)86.7
7MiMo-V2-Omni85.8
8Qwen 3 VL 235B A22B Instruct85.5
9Qwen 3 VL 4B Instruct85.5
10Qwen 3 VL 8B Instruct84.7
11Qwen 2.5 VL 32B Instruct84.7
12Qwen 3 VL 8B (Thinking)84.2
13GLM-4.5V83.1
14Qwen 3.5 4B79.6
15GPT-578.9

Interactive version: theaggregate.ai/benchmark?slug=chartarena-pie-chart-chinese · How It Works · Data refreshed daily, snapshot 2026-09-29.