ChartLlama: benchmark results

Provider: Other.

Aggregate Rating 1247 ± 51, rank #9137 of 10516 rated models, from 14 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ChartBench - Chart Recognition49.86Acc+ (%)60
ChartBench21.3Five-task average, Acc+ and NumberQA (%)55
ChartBench - NumberQA17.52Accuracy within 5% of the value (%)55
ChartBench - Value Comparison20.43Acc+ (%)50
ChartBench - Global Conception10.67Acc+ (%)45
LegendBench - Reasoning22.92Reasoning accuracy (%; raw accuracy on legend-conditioned qu36.4
ChartBench - Value Extraction8.38Acc+ (%)30
LegendBench - Guide Reading16.57Guide-reading accuracy (%; legend-parsing questions: extract27.3
LegendBench - Value Extraction6.98Value-extraction accuracy (%; legend-grounding questions: re27.3
LegendBench12.26QA accuracy (%; proportion of all individual questions answe18.2
LegendBench - Abstention F10Abstention F1 (%; harmonic mean of abstention precision and 18.2
CharXiv-R14.2Validation chart-visible reasoning accuracy (%)9.6

Interactive version: theaggregate.ai/model?slug=chartllama · How It Works · Data refreshed daily, snapshot 2026-11-21.