LongDS v1 - Business: leaderboard
Metric: Accuracy (%; macro-average over 12 business tasks of per-task turn accuracy; DSGym ReAct agent in a persistent Jupyter kernel, up to 40 steps per turn, DeepSeek-V4-Pro judge). Source: arxiv.org. Saturation forecast: Around October 2027. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 33.59 |
| 2 | GPT-5.4 | 28.4 |
| 3 | Claude Sonnet 4.6 | 25.54 |
| 4 | Kimi K2.6 | 20.99 |
| 5 | DeepSeek V4 Pro | 17.06 |
Interactive version: theaggregate.ai/benchmark?slug=longds-v1-business · How It Works · Data refreshed daily, snapshot 2026-09-25.