Kaggle RSNA Knee AUC10 - Vision (Large): leaderboard

Vision-only knee MRI label-extraction benchmark on Kaggle (kirderf/rsna-knee-llm-label-extraction-benchmark-auc10) measuring macro ROC-AUC (%) over 12 radiology findings across 10 radiologist-labeled consensus gold studies using large image mosaics.

Metric: Macro ROC-AUC (%). Source: www.kaggle.com. Saturation forecast: Around March 2027. 26 models tracked. Updated 21 Nov 2026.

Top models

#ModelScoreOverall rank
1GPT-6.1 Sol86.42#15
2GPT-6 Luna78.39#105
3GPT-5.6 Luna77.09#131
4GPT-5.575.74#54
5GPT-5.6 Sol67.66#37
6Claude Sonnet 5.566.41#24
7GPT-5.4 Mini66.31#406
8Gemini 3.1 Pro (Preview)65.69#99
9Claude Opus 5.564.13#9
10GPT-5.6 Terra62.92#71
11Gemini 3.8 Flash62.74#43
12Gemini 2.5 Pro61.79#276
13Gemini 4 Argon61.07#19
14Gemma 4 31B60.88#382
15Claude Haiku 4.559.81#583

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=kaggle-rsna-knee-auc10-vision-large · How It Works · Data refreshed daily, snapshot 2026-11-21.