DSGym - DSPredict Easy - Percentile: leaderboard
DSPredict-Easy, DSGym's split of Kaggle Playground Series and introductory competitions, scored on the private leaderboard and run by DSGym's default CodeAct agent (Nie et al., 2026) in an isolated Jupyter container at temperature 0 with all tools disabled. Mean leaderboard percentile of the submission (the share of other submissions it outscores), in percent; higher is better. The paper reports percentile here because medal rates saturate on these competitions.
Metric: Mean Leaderboard Percentile (%; private leaderboard). Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (High) | 60.4 | #225 (GPT-5.1) |
| 2 | GPT-5.1 (Medium) | 55.7 | #225 (GPT-5.1) |
| 3 | GPT-5 (Medium) | 53.5 | #163 (GPT-5) |
| 4 | Claude Sonnet 4.5 | 49 | #244 |
| 5 | GPT-5.1 (Non-reasoning) | 45.7 | #225 (GPT-5.1) |
| 6 | Claude Sonnet 4 | 44.4 | #397 |
| 7 | Kimi K2 0905 | 43.9 | #555 |
| 8 | DeepSeek V3.1 | 30.9 | #500 |
| 9 | Qwen 3 4B Instruct | 28.2 | #1660 |
| 10 | Qwen 2.5 7B Instruct | 17.5 | #2583 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dsgym-dspredict-easy-percentile · How It Works · Data refreshed daily, snapshot 2026-11-12.