MOSS-Audio Models by MOSI AI
Last refreshed 2026-06-29. Next refresh: weekly.
Details
Capabilities
About
MOSS-Audio is an open-weight audio-language model family for unified audio understanding across speech, environmental sound, music, captioning, time-aware question answering, timestamped transcription, and audio-grounded reasoning. The April 2026 release includes 4B and 8B Instruct and Thinking variants built with a dedicated audio encoder, modality adapter, and Qwen3 language-model backbones.
Decision facts
- Best fit
- multimodalaudio understandingvision and multimodal work
- Capability starting point
- MOSS-Audio 4B Thinking with reasoning and multimodal inputs
- Lowest tracked input
- Not tracked
- Closest related family
- MOVA
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs audio understanding, 4.6B parameters, and multimodal inputs.
Use when the workload needs audio understanding, 4.6B parameters, and reasoning.
Use when the workload needs audio understanding, 8.6B parameters, and multimodal inputs.
Use when the workload needs audio understanding, 8.6B parameters, and reasoning.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| MOSS-Audio 4B Instruct | Use when the workload needs audio understanding, 4.6B parameters, and multimodal inputs. | 2026-04 | audio understanding4.6B parametersmultimodal inputs | Current |
| MOSS-Audio 4B Thinking | Use when the workload needs audio understanding, 4.6B parameters, and reasoning. | 2026-04 | audio understanding4.6B parametersreasoning | Current |
| MOSS-Audio 8B Instruct | Use when the workload needs audio understanding, 8.6B parameters, and multimodal inputs. | 2026-04 | audio understanding8.6B parametersmultimodal inputs | Current |
| MOSS-Audio 8B Thinking | Use when the workload needs audio understanding, 8.6B parameters, and reasoning. | 2026-04 | audio understanding8.6B parametersreasoning | Current |
Release Timeline
1 release groupSpecifications(4 models)
| Model | Released | Parameters | Multimodal | Reasoning |
|---|---|---|---|---|
| MOSS-Audio 4B Instruct | 2026-04 | 4.6B | Yes | No |
| MOSS-Audio 4B Thinking | 2026-04 | 4.6B | Yes | Yes |
| MOSS-Audio 8B Instruct | 2026-04 | 8.6B | Yes | No |
| MOSS-Audio 8B Thinking | 2026-04 | 8.6B | Yes | Yes |
