Audio modelOpen weights
MOSS-Audio-4B-Thinking
OpenMOSS Team
- Released
- April 13, 2026
- Data date
- October 3, 2026
- Model class
- Access
MOSS-Audio-4B-Thinking combines audio understanding with text reasoning. The roughly 4.6-billion-parameter model accepts audio and text and can answer questions, caption recordings, or produce timestamped transcripts.
The output remains text. Despite its audio input, MOSS-Audio-4B-Thinking is not a speech synthesizer and does not produce a spoken answer. The weights are available locally under Apache-2.0.
Specifications and access
| Specification | Value and source |
|---|---|
| Model class | Audio understanding and text reasoningSource |
| Access | Local weightsSource |
| Input | Audio and textSource |
| Output | TextSource |
| Checkpoint | OpenMOSS-Team/MOSS-Audio-4B-ThinkingSource |
| License | Apache-2.0Source |
| Tasks | ASR, audio question answering, audio captioning, and timestamped ASRSource |
| Reasoning | Generates text answers with a Qwen3-4B backbone, not spoken audio responsesSource |
| Repository | Official OpenMOSS Audio repositorySource |
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
MMAU (Show measurement, test conditions, and source)
- Source value
- 77.64
- Score
- 77.64
- Metric
- score
- Unit
- No unit provided
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MOSS-Audio model card
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Exact 4B-Thinking row; audio understanding table, not ASR
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
MMAU-Pro (Show measurement, test conditions, and source)
- Source value
- 60.75
- Score
- 60.75
- Metric
- score
- Unit
- No unit provided
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MOSS-Audio model card
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Exact 4B-Thinking row; audio understanding table, not ASR
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
MMAR (Show measurement, test conditions, and source)
- Source value
- 63.91
- Score
- 63.91
- Metric
- score
- Unit
- No unit provided
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MOSS-Audio model card
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Exact 4B-Thinking row; audio understanding table, not ASR
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
MMSU (Show measurement, test conditions, and source)
- Source value
- 71.2
- Score
- 71.2
- Metric
- score
- Unit
- No unit provided
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MOSS-Audio model card
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Exact 4B-Thinking row; audio understanding table, not ASR
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
General Audio Understanding average accuracy (Show measurement, test conditions, and source)
- Source value
- 68.37
- Score
- 68.37
- Metric
- accuracy
- Unit
- No unit provided
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MOSS-Audio model card
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Exact 4B-Thinking row; audio understanding table, not ASR
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
Sources and data date
Every statement links to its underlying documentation or leaderboard.
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 4, 2026. 1 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Additional source | MOSS-Audio-4B-Thinking model card (retrieved October 3, 2026; October 4, 2026) · MOSS-Audio-4B-Thinking model card · Editorial description reviewed October 3, 2026 |
| Additional source | MOSS-Audio repository (retrieved October 3, 2026) · MOSS-Audio repository · Editorial description reviewed October 3, 2026 |