MedLoCoMo - Answerable Accuracy: leaderboard
Answer accuracy on MedLoCoMo (2026): the percent of answerable questions about a long multi-admission patient history that a Gemini 3 Flash Preview judge marks correct, over both question scopes. Higher is better.
Metric: LLM-judge accuracy (%) on the answerable questions of both scopes; full patient history in context (100 synthetic multi-admission doctor-patient conversations from MIMIC-IV, 74.5k tokens on average), short free-text answers; answerable items judged correct or not by a Gemini 3 Flash Preview judge, adversarial unanswerable items scored by abstention; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 64.2 | #229 |
| 2 | Qwen 3.5 9B | 58.4 | #803 |
| 3 | Qwen 3.5 27B | 56.1 | #359 |
| 4 | Qwen 3.5 4B | 50.8 | #1175 |
| 5 | Gemma 3 27B | 33.9 | #1631 |
| 6 | Gemma 3 12B | 33.3 | #2000 |
| 7 | MedGemma-4B | 22.7 | #2444 |
| 8 | Gemma 3 4B | 18.6 | #5064 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medlocomo-answerable-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-13.