Spider 2.0-Snow: leaderboard
Text-to-SQL on real enterprise Snowflake data warehouses with 547 problems, translating natural language into correct SQL against large, complex schemas with documentation.
Metric: Accuracy (%). Source: spider2-sql.github.io. Saturation forecast: Estimated already saturated. 52 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 94.15 | #66 |
| 2 | Claude Sonnet 4.5 | 90.49 | #142 |
| 3 | Claude Opus 4.6 | 86.28 | #63 |
| 4 | DeepSeek V3.2 | 70.38 | #205 |
| 5 | GPT-5 | 65.63 | #98 |
| 6 | GPT-OSS-120B | 65.45 | #341 |
| 7 | DeepSeek R1 | 63.8 | #251 |
| 8 | O3 | 62.89 | #124 |
| 9 | Claude Sonnet 4 | 61.43 | #199 |
| 10 | Kimi K2.5 | 55.94 | #144 |
| 11 | GPT-5.4 Mini | 50.46 | #212 |
| 12 | GLM-5 | 50.27 | #141 |
| 13 | Qwen 3 Max | 45.15 | #208 |
| 14 | GPT-5 Mini | 40.95 | #186 |
| 15 | DeepSeek V3 | 38.03 | #321 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=spider-2-0-snow · How It Works · Data refreshed daily, snapshot 2026-10-20.