MTBBench-Multimodal - Hematology: leaderboard
Hematology questions of the MTBBench multimodal track (Vasilev et al., 2025): from a head-and-neck cancer patient's preoperative blood tests and reference ranges, the agent judges infection risk, bleeding tendency, renal impairment and similar findings. Accuracy in percent, no tools; higher is better.
Metric: Accuracy (%; preoperative hematology questions, 4 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 | InternVL3-38B | 79.7 | #905 |
| 2 | InternVL3-78B | 79.7 | #733 |
| 3 | O4 Mini (2025-04-16) | 77.8 | #362 |
| 4 | GPT-4o (2024-08-06) | 76.9 | #731 |
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-hematology · How It Works · Data refreshed daily, snapshot 2026-11-12.