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

#ModelScore
1DeepSeek V389.06
2Qwen 3 235B A22B 250788.03
3Qwen 3 32B85.77
4Qwen 3 14B85.43
5Qwen 3 235B A22B84.54
6DeepSeek R1 Distill Qwen 14B83.01
7Qwen 3 30B A3B82.67
8DeepSeek R1 Distill Qwen 32B82.29
9Qwen 3 1.7B69.04
10Qwen 3 8B68.5
11DeepSeek-R1-Distill-Qwen-7B67.13
12Qwen 3 4B64.29

Interactive version: theaggregate.ai/benchmark?slug=medrepbench-abnormal-flag-recall-ocr-text-only · How It Works · Data refreshed daily, snapshot 2026-09-26.