RealDataAgentBench — leaderboard
Data-science agent benchmark evaluating whether LLM agents solve real-data analysis tasks correctly and robustly across correctness, code quality, efficiency, and statistical validity.
Metric: Average DAB Score (%). Source: patibandlavenkatamanideep.github.io. Status: saturation imminent. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 | 88.89 |
| 2 | GPT-4.1 | 87.52 |
| 3 | GPT-4.1 Mini | 87.02 |
| 4 | Claude Sonnet 4.6 | 85.74 |
| 5 | GPT-4o | 85.13 |
| 6 | Claude Opus 4.6 | 84.62 |
| 7 | Grok 3 Mini | 82.69 |
| 8 | Claude Haiku 4.5 (20251001) | 80.05 |
| 9 | GPT-4o Mini | 78.49 |
| 10 | GPT-5 | 77.95 |
| 11 | Gemini 2.5 Flash | 66.21 |
| 12 | GPT-4.1 Nano | 62.37 |
Interactive version: theaggregate.ai/benchmark?slug=realdataagentbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.