LiveBench's data analysis category — tests the ability to interpret, analyze, and draw conclusions from structured data.
Covers structured-data interpretation including tables, charts, and quantitative comparisons; verifiable outputs reduce grader subjectivity for analytic tasks.
Tasks use cleaner inputs than most production datasets, so real-world ETL noise is underrepresented; benchmark scope is narrower than full analysis workflows with data cleaning and tooling.
Higher is better. Accuracy on data analysis tasks.
Max score: 100
Tasks use cleaner inputs than most production datasets, so real-world ETL noise is underrepresented.
| # | Model | Score |
|---|---|---|
| 1 | 97.6 | |
| 2 | 96.0 | |
| 3 | 94.4 | |
| 4 | 86.5 | |
| 5 | 85.3 | |
| 6 | 85.2 | |
| 7 | 84.1 | |
| 8 | 83.4 | |
| 9 | 83.2 | |
| 10 | 83.0 | |
| 11 | 83.0 | |
| 12 | 82.9 | |
| 13 | 82.7 | |
| 14 | 82.3 | |
| 15 | 81.6 | |
| 16 | 81.2 | |
| 17 | 81.1 | |
| 18 | 80.8 | |
| 19 | 80.5 | |
| 20 | 80.3 | |
| 21 | 80.3 | |
| 22 | 80.2 | |
| 23 | 79.8 | |
| 24 | 79.6 | |
| 25 | 79.6 | |
| 26 | 79.5 | |
| 27 | 79.5 | |
| 28 | 79.5 | |
| 29 | 79.5 | |
| 30 | 79.3 | |
| 31 | 79.3 | |
| 32 | 79.3 | |
| 33 | 79.3 | |
| 34 | 79.2 | |
| 35 | 79.2 | |
| 36 | 79.2 | |
| 37 | 79.1 | |
| 38 | 78.7 | |
| 39 | 78.5 | |
| 40 | 78.5 | |
| 41 | 78.4 | |
| 42 | 78.4 | |
| 43 | 78.2 | |
| 44 | 78.0 | |
| 45 | 77.9 | |
| 46 | 77.9 | |
| 47 | 77.3 | |
| 48 | 77.3 | |
| 49 | 77.2 | |
| 50 | 77.0 |