VAREX-Bench: leaderboard
Multi-modal structured extraction benchmark over 1,777 government forms with per-document JSON schemas and image/text/spatial modalities.
Metric: Exact Match (imported). Source: huggingface.co. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 98 | #152 |
| 2 | Gemini 2.5 Flash | 97.3 | #245 |
| 3 | Llama 4 Maverick | 95.6 | #464 |
| 4 | GPT-4o | 94.8 | #343 |
| 5 | Llama 4 Scout | 94.3 | #656 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=varex-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.