MMBU - Grounded Classification (Masks, Open-Ended): leaderboard
Metric: Open-ended classification of a region marked by a segmentation mask: micro-averaged F1 x100 (0-100) over MMBU's biomedical images from 35 submodalities (95% bootstrap intervals not published); zero-shot, official checkpoints; Qwen3-32B judges strict biomedical equivalence. Source: arxiv.org. Saturation forecast: Around September 2027. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 32B Instruct | 18.2 |
| 2 | GPT-5.4 Mini | 12.6 |
| 3 | Qwen 2.5 VL 32B Instruct | 11.5 |
| 4 | InternVL3.5-8B | 8.8 |
| 5 | Qwen 3 VL 8B Instruct | 7.4 |
| 6 | GPT-4.1 Mini | 6.5 |
| 7 | Qwen 3 VL 4B Instruct | 6.5 |
| 8 | Lingshu-32B | 5.3 |
| 9 | Lingshu-7B | 4.9 |
| 10 | MedGemma-4B-IT | 4.2 |
| 11 | Qwen 2.5 VL 7B Instruct | 3 |
| 12 | Gemma 3 4B (IT) | 0.9 |
Interactive version: theaggregate.ai/benchmark?slug=mmbu-grounded-classification-masks-open-ended · How It Works · Data refreshed daily, snapshot 2026-09-29.