ChartArena - Line Chart (English): leaderboard

Metric: mAP-high (%) on English line 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: Around December 2026. 44 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)67
2Kimi K2.5 (Non-reasoning)60.9
3Qwen 3.5 35B A3B (Thinking)60.3
4Gemini 2.5 Pro56.5
5Seed 2.0 Pro (Non-reasoning)56.5
6Gemma 4 31B (IT)53.4
7Qwen 3 VL 235B A22B Instruct52.3
8GLM-4.5V51.7
9GPT-548.1
10Qwen 3.5 9B45.5
11Qwen 3 VL 8B (Thinking)42
12MiMo-V2-Omni41.5
13Qwen 2.5 VL 72B Instruct38.2
14Qwen 3 VL 8B Instruct35.5
15Qwen 3.5 4B35.1

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