TableBench: leaderboard
TableBench: Measures structured-data reasoning over tables, spreadsheets, charts, databases, or data analysis tasks.
Metric: DP Score (%). Source: github.com. Source: tablebench.github.io. Saturation forecast: Around January 2027. 32 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (High) | 61.69 | #180 (O4 Mini) |
| 2 | O4 Mini | 60.75 | #180 |
| 3 | GPT-5 | 59.94 | #98 |
| 4 | O3 Mini | 59.9 | #272 |
| 5 | Grok 4 | 57.8 | #181 |
| 6 | Gemini 2.5 Pro | 57.18 | #152 |
| 7 | DeepSeek R1 | 56.31 | #251 |
| 8 | Claude Sonnet 4 | 54.75 | #199 |
| 9 | Qwen 3 32B | 52.45 | #443 |
| 10 | Llama 4 Maverick Instruct | 49.49 | #456 |
| 11 | QwQ-32B | 43.87 | #421 |
| 12 | GPT-4o | 40.91 | #343 |
| 13 | GPT-4 Turbo | 38.74 | #404 |
| 14 | DeepSeek V3 | 36.99 | #321 |
| 15 | Llama 3.1 405B Instruct | 36.04 | #457 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=tablebench · How It Works · Data refreshed daily, snapshot 2026-10-20.