DARE-bench (Data Science) - Time-Series Forecasting (Canonical): leaderboard
Metric: Clipped R-squared (0 to 1, shown times 100) on the 57 canonical forecasting tasks whose test inputs keep only the timestamp and entity columns; DARE-bench test tasks derived from recently updated Kaggle datasets; the model works as a data-science agent with a sandboxed Python execution tool (5 interaction turns, 200 s per execution, greedy decoding), mean of three repeats; higher is better. Source: arxiv.org. Saturation forecast: Around June 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 13.7 | #241 |
| 2 | GPT-5 | 10.13 | #91 |
| 3 | O4 Mini | 9.67 | #172 |
| 4 | GPT-4.1 | 6.6 | #240 |
| 5 | GPT-4o | 4.77 | #333 |
| 6 | Claude Sonnet 4 | 0.01 | #194 |
| 7 | Qwen 3 32B | 0 | #424 |
| 8 | Qwen 3 4B | 0 | #823 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=dare-bench-data-science-time-series-forecasting-canonical · How It Works · Data refreshed daily, snapshot 2026-10-11.