Kaggle ExtractBench: leaderboard
LlamaIndex's ExtractBench leaderboard on Kaggle: schema-guided extraction of structured data from enterprise documents (370 documents, 4,869 pages and 8 business domains in the paper). This board is Kaggle's score on the full set, extractbench-full, converted from a 0-1 fraction to 0-100, higher is better.
Metric: Score (%). Source: www.kaggle.com. Saturation forecast: Around February 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol | 91.04 | #22 |
| 2 | GPT-6 Astra | 90.28 | #6 |
| 3 | GPT-5.6 Terra | 89.99 | #40 |
| 4 | Gemini 4 Argon | 89.81 | #9 |
| 5 | GPT-5.5 | 89.11 | #33 |
| 6 | Claude Sonnet 5.5 | 89.01 | #14 |
| 7 | Gemini 3 Flash (Preview) | 88.98 | #90 |
| 8 | GPT-5.6 Luna | 88.97 | #79 |
| 9 | Claude Opus 5 | 88.81 | #16 |
| 10 | Claude Sonnet 5 | 88.55 | #65 |
| 11 | Gemini 3.1 Pro (Preview) | 87.64 | #64 |
| 12 | Gemini 3.8 Flash | 87.15 | #30 |
| 13 | GPT-5.4 Mini | 86.35 | #228 |
| 14 | Gemini 3.5 Flash | 85.56 | #66 |
| 15 | Gemini 3.7 Flash | 85.47 | #24 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-extractbench · How It Works · Data refreshed daily, snapshot 2026-11-05.