| license: apache-2.0 | |
| language: | |
| - en | |
| - zh | |
| - ja | |
| - ko | |
| - fr | |
| - de | |
| - es | |
| tags: | |
| - speech | |
| - asr | |
| - mlx | |
| - qwen3 | |
| library_name: mlx | |
| base_model: Qwen/Qwen3-ASR-0.6B | |
| # Qwen3 ASR 0.6B — MLX 4-bit | |
| MLX 4-bit quantized conversion of [Qwen/Qwen3-ASR-0.6B](https://huggingface.co/Qwen/Qwen3-ASR-0.6B) for Apple Silicon inference. | |
| ## Usage | |
| Used by [qwen3-asr-swift](https://github.com/AufklarerStudios/qwen3-asr-swift) `Qwen3ASR` module: | |
| ```swift | |
| let model = try await Qwen3ASRModel.fromPretrained() | |
| let text = model.transcribe(audio: samples, sampleRate: 16000) | |
| ``` | |
| ```bash | |
| audio transcribe audio.wav | |
| ``` | |
| ## Model Details | |
| - **Architecture**: Qwen3-ASR encoder-decoder (Whisper-style audio encoder + Qwen3 text decoder) | |
| - **Parameters**: 0.6B | |
| - **Quantization**: 4-bit (MLX) | |
| - **Size**: ~680 MB | |
| - **Languages**: Multilingual (EN, ZH, JA, KO, FR, DE, ES, and more) | |