MedMT-Bench - Contextual Interference: leaderboard
Metric: Accuracy (%) on the 41 resistance-to-contextual-interference dialogues (injected history in a mismatched style or format) of MedMT-Bench's long multi-turn medical dialogues (health consultation, treatment recommendation and rehabilitation scenarios, text and image-text), where the final-turn reply is checked against fine-grained test points, judged by human annotators; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (High) | 60.98 | #91 (GPT-5) |
| 2 | Gemini 2.5 Pro (Preview 06-05) | 60.98 | #101 |
| 3 | GPT-4.1 | 46.34 | #240 |
| 4 | Claude Sonnet 4 | 46.34 | #194 |
| 5 | O3 (2025-04-16) | 39.02 | #117 |
| 6 | Claude Opus 4 | 39.02 | #155 |
| 7 | GPT-4o (2024-11-20) | 21.95 | #369 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=medmt-bench-contextual-interference · How It Works · Data refreshed daily, snapshot 2026-10-11.