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

#ModelScore
1Gemini 3.1 Pro (Preview)33.59
2GPT-5.428.4
3Claude Sonnet 4.625.54
4Kimi K2.620.99
5DeepSeek V4 Pro17.06

Interactive version: theaggregate.ai/benchmark?slug=longds-v1-business · How It Works · Data refreshed daily, snapshot 2026-09-25.