BlueLM-2.5-3B: benchmark results
Provider: vivo. Released 2025-07-01.
Aggregate Rating 1581 ± 29, rank #1235 of 10916 rated models, from 154 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SuperCLUE-Ondevice (September 2025) - Multimodal | 85.02 | Score | 100 |
| OpenVLM MathVista - MWP | 93.5 | Accuracy (%) | 99.7 |
| OpenVLM MathVista - ARI | 74.5 | Accuracy (%) | 98.8 |
| OpenVLM HallusionBench - qAcc | 54.1 | Accuracy (%) | 97.2 |
| OpenVLM MMBench V1.1 EN - Image Quality | 68.5 | Accuracy (%) | 97.2 |
| OpenVLM MMMU - History | 83.3 | Accuracy (%) | 96.5 |
| OpenVLM MMMU - Math | 66.7 | Accuracy (%) | 95.9 |
| OpenVLM HallusionBench - aAcc | 73.7 | Accuracy (%) | 95.4 |
| OpenVLM MathVista - STA | 85.7 | Accuracy (%) | 94.9 |
| OpenVLM MMStar - Math | 82 | Accuracy (%) | 94.7 |
| OpenVLM MMStar - Logical Reasoning | 77.6 | Accuracy (%) | 94.5 |
| OpenVLM MMMU - Finance | 80 | Accuracy (%) | 94.3 |
Interactive version: theaggregate.ai/model?slug=bluelm-2-5-3b · How It Works · Data refreshed daily, snapshot 2026-11-08.