ProLLM - SQL Disambiguation — leaderboard
ProLLM benchmark measuring LLM ability to disambiguate and generate correct SQL from ambiguous natural language queries.
Metric: Score (%). Source: www.prollm.ai. Status: saturation imminent. 52 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.5 | 53.1 |
| 2 | O1 | 49.2 |
| 3 | Grok 2 | 47.9 |
| 4 | GPT-3.5 Turbo | 47.7 |
| 5 | GPT-4.1 | 46.4 |
| 6 | Qwen 2.5 72B Instruct | 46 |
| 7 | DeepSeek V3 | 44.7 |
| 8 | Gemini 2.0 Flash | 42 |
| 9 | Mistral Small 3.1 | 40.8 |
| 10 | Command A | 40.6 |
| 11 | Grok 3 | 40.3 |
| 12 | GPT-4.1 Mini | 40.1 |
| 13 | GPT-4 Turbo | 37.5 |
| 14 | Nova Pro | 37.4 |
| 15 | GPT-4o | 36.7 |
Interactive version: theaggregate.ai/benchmark?slug=prollm-sql-disambiguation · How the rankings work · Data refreshed daily, snapshot 2026-07-22.