DBLifeBench - Design: leaderboard
The design phase of DBLifeBench: from natural-language requirements a model designs a two-table relational schema, scored as the mean of entity-name, data-type and foreign-key accuracy in percent; higher is better.
Metric: Schema design accuracy (%; mean of entity-name, data-type and foreign-key accuracy of two-table schemas designed from 265 natural-language requirements; temperature 0.3, top-p 0.2, SQLite). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 74.63 | #741 |
| 2 | GPT-4o Mini | 68.36 | #1623 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dblifebench-design · How It Works · Data refreshed daily, snapshot 2026-11-13.