ChartLlama: benchmark results
Provider: Other.
Aggregate Rating 1247 ± 51, rank #9137 of 10516 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ChartBench - Chart Recognition | 49.86 | Acc+ (%) | 60 |
| ChartBench | 21.3 | Five-task average, Acc+ and NumberQA (%) | 55 |
| ChartBench - NumberQA | 17.52 | Accuracy within 5% of the value (%) | 55 |
| ChartBench - Value Comparison | 20.43 | Acc+ (%) | 50 |
| ChartBench - Global Conception | 10.67 | Acc+ (%) | 45 |
| LegendBench - Reasoning | 22.92 | Reasoning accuracy (%; raw accuracy on legend-conditioned qu | 36.4 |
| ChartBench - Value Extraction | 8.38 | Acc+ (%) | 30 |
| LegendBench - Guide Reading | 16.57 | Guide-reading accuracy (%; legend-parsing questions: extract | 27.3 |
| LegendBench - Value Extraction | 6.98 | Value-extraction accuracy (%; legend-grounding questions: re | 27.3 |
| LegendBench | 12.26 | QA accuracy (%; proportion of all individual questions answe | 18.2 |
| LegendBench - Abstention F1 | 0 | Abstention F1 (%; harmonic mean of abstention precision and | 18.2 |
| CharXiv-R | 14.2 | Validation chart-visible reasoning accuracy (%) | 9.6 |
Interactive version: theaggregate.ai/model?slug=chartllama · How It Works · Data refreshed daily, snapshot 2026-11-21.