MedRepBench - Abnormal Flag Recall (OCR Text Only): leaderboard
Metric: Field-level recall (%; abnormality flag of each ground-truth lab item, matched by exact normalized item name within the model's top-K items, K = number of reference items; text LLM on the OCR text). Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V3 | 89.06 |
| 2 | Qwen 3 235B A22B 2507 | 88.03 |
| 3 | Qwen 3 32B | 85.77 |
| 4 | Qwen 3 14B | 85.43 |
| 5 | Qwen 3 235B A22B | 84.54 |
| 6 | DeepSeek R1 Distill Qwen 14B | 83.01 |
| 7 | Qwen 3 30B A3B | 82.67 |
| 8 | DeepSeek R1 Distill Qwen 32B | 82.29 |
| 9 | Qwen 3 1.7B | 69.04 |
| 10 | Qwen 3 8B | 68.5 |
| 11 | DeepSeek-R1-Distill-Qwen-7B | 67.13 |
| 12 | Qwen 3 4B | 64.29 |
Interactive version: theaggregate.ai/benchmark?slug=medrepbench-abnormal-flag-recall-ocr-text-only · How It Works · Data refreshed daily, snapshot 2026-09-26.