MedQ-Deg - Severe Degradation (L2): leaderboard
Metric: Accuracy (%) on the MedQ-Deg items with severe degradations (L2, diagnosis difficult but still feasible), multiple-choice medical VQA items merged from OmniMedVQA, GMAI-MMBench and MedXpertQA, each image corrupted by modality-specific degradations whose severity three radiologists calibrated (24,894 QA pairs over 7 modalities and 18 degradation types), single-letter answer, temperature 1.0, mean of 3 runs; higher is better. Source: arxiv.org. Saturation forecast: Around March 2028. 40 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 66.4 | #264 |
| 2 | Qwen 3 VL 30B A3B Instruct | 66.14 | #365 |
| 3 | Gemini 2.5 Pro | 65.2 | #145 |
| 4 | GPT-5 | 64.3 | #91 |
| 5 | GPT-5.1 | 62.16 | #131 |
| 6 | Gemini 2.5 Flash | 62 | #237 |
| 7 | GPT-4.1 | 62 | #240 |
| 8 | GPT-4.1 Mini | 61.67 | #346 |
| 9 | GPT-4o | 59.73 | #333 |
| 10 | GLM-4.5V | 58.94 | #339 |
| 11 | MedGemma-4B | 53.77 | #731 |
| 12 | Qwen 2.5 VL 72B Instruct | 52.92 | #364 |
| 13 | Claude Sonnet 4.5 | 52.86 | #138 |
| 14 | Gemma 3 27B | 52.45 | #596 |
| 15 | GPT-4o Mini | 51.53 | #588 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=medq-deg-severe-degradation-l2 · How It Works · Data refreshed daily, snapshot 2026-10-11.