MedicalMLOmni - Medical VQA Accuracy: leaderboard
MedicalMLOmni (2026): fixed evidence-planning and parallel-HPO framework, four language-model backbones; resulting medical model quality over8segmentation and7VQA tasks. These are AutoML systems, not direct medical answers from the backbone.18GPU-hour tuning budget; framework ablations and task-specific human/classic baselines kept out of this backbone comparison. Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | MedicalMLOmni System GPT-4o | 69.5 | |
| 2 | MedicalMLOmni System Llama-3.3-70B | 68.2 | |
| 3 | MedicalMLOmni System Qwen3-30B | 67.7 | |
| 4 | MedicalMLOmni System Qwen3-8B | 62.1 |
Interactive version: theaggregate.ai/benchmark?slug=medicalmlomni-medical-vqa-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-22.