MOSS-Audio-8B: benchmark results
Provider: Other.
Aggregate Rating 1521 ± 50, rank #2055 of 10916 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VoiceGiraffe (Cascaded Captions) | 45.3 | Overall accuracy (%) over the 1,500 four-option VoiceGiraffe | 54.5 |
| VoiceGiraffe (Cascaded Captions) - Single-Hop Perception | 44.8 | Accuracy (%) on the 1,000 single-hop questions (temporal loc | 54.5 |
| VoiceGiraffe (Cascaded Captions) - Multi-Hop Reasoning | 46.4 | Accuracy (%) on the 500 multi-hop questions (causal alignmen | 40.9 |
| MuLA-Bench - Acoustic Evidence | 13.11 | Accuracy (%; the 1,198 acoustic-track questions about natura | 22.2 |
| MuLA-Bench - Chinese | 25.8 | Accuracy (%; the 337 Chinese-language questions over both ev | 14.3 |
| MuLA-Bench - Reasoning | 14.49 | Accuracy (%; the 1,146 reasoning questions requiring inferen | 11.1 |
| MuLA-Bench - Temporal Localization | 6.75 | Accuracy (%; the 1,008 temporal questions on event order and | 11.1 |
| MuLA-Bench | 13.48 | Accuracy (%; all 5,038 questions over 1,769 recordings in 16 | 0 |
| MuLA-Bench - Arabic | 3.5 | Accuracy (%; the 313 Arabic-language questions over both evi | 0 |
| MuLA-Bench - Belief Domain | 11.5 | Accuracy (%; the 480 semantic-track questions from belief an | 0 |
| MuLA-Bench - English | 25.4 | Accuracy (%; the 347 English-language questions over both ev | 0 |
| MuLA-Bench - Entertainment Domain | 15.8 | Accuracy (%; the 480 semantic-track questions from entertain | 0 |
Interactive version: theaggregate.ai/model?slug=moss-audio-8b · How It Works · Data refreshed daily, snapshot 2026-11-08.