MeniOmni - Report Generation: leaderboard

MeniOmni (2026) radiology report generation: from three knee MRI sequences alone, a multimodal model writes the findings and impression, scored by the Meni-Score, an F1 over canonicalized clinical entities (location, morphology, signal) shared with the expert report, in percent; higher is better.

Metric: Meni-Score (%): F1 between the clinical entities (location, morphology, signal) extracted and canonicalized from the generated report and from the expert report, zero-shot radiology report generation from the three MRI sequences alone (no demographics), temperature 0.3, on the MeniOmni test split; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4o52.4#751
2Claude 3.5 Sonnet51#752
3Gemini 1.5 Pro48.5#1063
4Qwen 2 VL 72B46.5#1134
5MeniOmni InternVL2 (checkpoint unspecified)44.2
6Video-LLaVA-7B32.4#3879

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

Interactive version: theaggregate.ai/benchmark?slug=meniomni-report-generation · How It Works · Data refreshed daily, snapshot 2026-11-21.