ChartArena - Chinese: leaderboard
ChartArena (2026) tests chart parsing: a model converts a chart image into structured data, scored format-agnostically as triples for numeric charts or as labeled graphs and trees for flowcharts and mind maps. This board is mAP at the high tolerance level in percent, averaged over eight chart types and three visual styles (digital, printed photo, hand-drawn) on Chinese charts; higher is better.
Metric: mAP-high (%) averaged over the eight chart types (bar, line, pie, radar, box plot, combination, flowchart, mind map) on Chinese 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 February 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 73.2 | #94 |
| 2 | Kimi K2.5 (Non-reasoning) | 68.1 | #256 (Kimi K2.5) |
| 3 | Qwen 3.5 35B A3B (Thinking) | 65.5 | #482 (Qwen 3.5 35B A3B) |
| 4 | Gemini 2.5 Pro | 62.4 | #266 |
| 5 | Qwen 3 VL 235B A22B Instruct | 58.4 | #522 |
| 6 | Gemma 4 31B (IT) | 57.6 | #342 |
| 7 | GLM-4.5V | 53.9 | #744 |
| 8 | Qwen 3 VL 8B Instruct | 50.4 | #981 |
| 9 | Qwen 2.5 VL 72B Instruct | 50 | #846 |
| 10 | Qwen 3 VL 8B (Thinking) | 49.7 | #887 |
| 11 | Qwen 3.5 9B | 47.7 | #803 |
| 12 | Qwen 2.5 VL 32B Instruct | 47.7 | #1053 |
| 13 | GPT-5 | 45.8 | #167 |
| 14 | Qwen 3 VL 4B Instruct | 45.6 | #1279 |
| 15 | Qwen 3 VL 4B (Thinking) | 44.3 | #1296 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chartarena-chinese · How It Works · Data refreshed daily, snapshot 2026-11-13.