TinyChart: benchmark results
Provider: Other.
Aggregate Rating 1133 ± 46, rank #10264 of 10916 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Self-Ensembling Vision-Language Models for Cha | 95.2 | ChartQA (self-reported) | 92.3 |
| ThermEval - Thermal Reasoning (One Person) | 36 | Accuracy (times 100, 0-100) on Task 5, ranking one person's | 54.3 |
| LegendBench - Value Extraction | 18.3 | Value-extraction accuracy (%; legend-grounding questions: re | 36.4 |
| ThermEval - Temperature Estimation (Coordinates) | 6.68 | Mean absolute error (degrees Celsius, 0 or more) on Task 6, | 34.8 |
| LegendBench | 13.77 | QA accuracy (%; proportion of all individual questions answe | 27.3 |
| LegendBench - Binding | 6.32 | Binding accuracy (%; legend-grounding questions: associate a | 27.3 |
| EpiCurveBench | 11.4 | EpiCurveSimilarity, ECS (%): edit distance with real penalty | 20 |
| LegendBench - Abstention F1 | 0 | Abstention F1 (%; harmonic mean of abstention precision and | 18.2 |
| LegendBench - Guide Reading | 16.53 | Guide-reading accuracy (%; legend-parsing questions: extract | 18.2 |
| LegendBench - Reasoning | 15.53 | Reasoning accuracy (%; raw accuracy on legend-conditioned qu | 18.2 |
| ThermEval - Temperature Estimation (Marker) | 5.66 | Mean absolute error (degrees Celsius, 0 or more) on Task 6, | 17.4 |
| ThermEval - Thermal Reasoning (Two People) | 39 | Accuracy (times 100, 0-100) on Task 5, whether the chest, fo | 17.4 |
Interactive version: theaggregate.ai/model?slug=tinychart · How It Works · Data refreshed daily, snapshot 2026-11-08.