DeSTA2.5-Audio: benchmark results
Provider: Other.
Aggregate Rating 1540 ± 45, rank #1840 of 10916 rated models, from 19 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MedMosaic - Speech and Sound MCQ | 41.2 | Accuracy (%; multiple choice, extracted answer letter agains | 83.3 |
| MedMosaic - Lung Sound MCQ | 44.6 | Accuracy (%; multiple choice, extracted answer letter agains | 75 |
| RAIL - Acquired Knowledge | 62.52 | LLM-as-judge score (%) on acquired-knowledge tasks over spee | 68 |
| RAIL - Auditory Processing | 46.75 | LLM-as-judge score (%) on auditory-processing tasks (speech- | 68 |
| MedMosaic | 41 | Weighted accuracy (%; category scores weighted by question c | 66.7 |
| MedMosaic - Speech MCQ | 49.4 | Accuracy (%; multiple choice, extracted answer letter agains | 66.7 |
| RAIL - Fluid Reasoning | 51.55 | LLM-as-judge score (%) on fluid-reasoning tasks (induction, | 60 |
| MedMosaic - Heart Sound MCQ | 37.1 | Accuracy (%; multiple choice, extracted answer letter agains | 58.3 |
| MMAU | 65.21 | Avg Test (imported) | 56.7 |
| MedMosaic - Long-Form Consultation MCQ | 24.9 | Accuracy (%; multiple choice, extracted answer letter agains | 50 |
| MedMosaic - Open-Ended Speech | 56 | Judge score (x 100; GPT-5.1 rates correctness, relevance, co | 50 |
| MedMosaic - Multi-Turn MCQ | 39.4 | Accuracy (%; multiple choice, extracted answer letter agains | 33.3 |
Interactive version: theaggregate.ai/model?slug=desta2-5-audio · How It Works · Data refreshed daily, snapshot 2026-11-08.