MMAlaya: benchmark results
Provider: DataCanvas. Released 2024-03-08. Access: Open.
Aggregate Rating 1239 ± 45, rank #9348 of 10916 rated models, from 212 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MathVista - LOG | 32.4 | Accuracy (%) | 85.7 |
| OpenVLM COCO Captions | 6.4 | CIDEr | 77.4 |
| OpenVLM MMMU - Music | 36.7 | Accuracy (%) | 75.4 |
| OpenVLM MMMU - Diagnostics and Laboratory Medicine | 40 | Accuracy (%) | 64.8 |
| OpenVLM MME - Scene | 158.8 | Score | 60 |
| OpenVLM MMBench V1.1 CN - Physical Property Reasoning | 60.8 | Accuracy (%) | 53.4 |
| OpenVLM OCR-VQA | 51 | Accuracy (%) | 42.3 |
| OpenVLM MMBench V1.1 CN - Action Recognition | 82.9 | Accuracy (%) | 40.8 |
| OpenVLM MMBench V1.1 CN - Identity Reasoning | 94.7 | Accuracy (%) | 37.1 |
| OpenVLM MMBench V1.1 CN - Image Emotion | 75.6 | Accuracy (%) | 35.9 |
| OpenVLM MMMU - Basic Medical Science | 46.7 | Accuracy (%) | 35.7 |
| OpenVLM MMBench V1.1 EN - Image Scene | 85.4 | Accuracy (%) | 33 |
Interactive version: theaggregate.ai/model?slug=mmalaya · How It Works · Data refreshed daily, snapshot 2026-11-08.