MeniOmni - Severe Error Rate: leaderboard
Metric: Severe error rate (SER, %): share of cases where either compartment is graded two or more grades away from the radiologists, zero-shot Stoller grading (0 to 3) of the medial and lateral meniscus from sagittal, coronal and axial knee MRI sequences plus serialized age, sex and BMI, on the held-out test split (about a tenth of 746 cases from three centres, patient-level split), temperature 0; lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 12.4 |
| 2 | Claude 3.5 Sonnet | 13.8 |
| 3 | Gemini 1.5 Pro | 15.2 |
| 4 | Qwen 2 VL 72B | 17.8 |
Interactive version: theaggregate.ai/benchmark?slug=meniomni-severe-error-rate · How It Works · Data refreshed daily, snapshot 2026-10-07.