DTBench: leaderboard
Metric: Cell-level F1 (%) of the table the model extracts from each of 120 long synthesized documents (498 to 182,297 tokens) against the ground-truth table (8,811 cells), after normalization, exact matching and bipartite row alignment; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 89.34 | #93 |
| 2 | GPT-5 | 85.99 | #91 |
| 3 | DeepSeek V3.2 | 85.3 | #198 |
| 4 | GPT-5 Mini | 84.47 | #176 |
| 5 | Qwen 3 32B | 62.98 | #424 |
| 6 | Qwen 3 4B | 33.65 | #823 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=dtbench · How It Works · Data refreshed daily, snapshot 2026-10-11.