MTBBench-Multimodal - Digital Pathology: leaderboard
Digital-pathology questions of the MTBBench multimodal track (Vasilev et al., 2025): the agent requests H&E and IHC images of a head-and-neck cancer patient and answers questions on histologic subtype and immune-cell patterns in the tumor. Accuracy in percent over true/false and six-option multiple-choice questions, no tools; higher is better.
Metric: Accuracy (%; H&E, IHC and joint H&E+IHC questions, 9 per patient; multi-turn file access, no tools). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-08-06) | 63.2 | #731 |
| 2 | InternVL3-78B | 62 | #733 |
| 3 | O4 Mini (2025-04-16) | 59.5 | #362 |
| 4 | InternVL3-38B | 54.7 | #905 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mtbbench-multimodal-digital-pathology · How It Works · Data refreshed daily, snapshot 2026-11-12.