PARROT - Result Consistency: leaderboard
The "Result Consistency" subtask of PARROT. PARROT SQL-dialect benchmark (code4db) — tests whether models translate and execute SQL correctly across database dialects, scoring dialect compatibility (including MySQL and Oracle) and cross-dialect result consistency.
Metric: Accuracy (%). Source: code4db.github.io. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini | 54.23 | #272 |
| 2 | O1 Preview | 48.69 | #193 |
| 3 | Doubao-1.5-Thinking-Pro | 25.7 | #134 |
| 4 | Claude 3.7 Sonnet | 22.74 | #249 |
| 5 | GPT-4o | 21.87 | #343 |
| 6 | DeepSeek R1 Distill Qwen 32B | 16.91 | #655 |
| 7 | Doubao-1.5-Pro | 14.29 | #287 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=parrot-result-consistency · How It Works · Data refreshed daily, snapshot 2026-10-20.