aitytech/Qwen3-ASR-1.7B-MLX-8bit

Qwen3 ASR 1.7B — MLX 8-bit

MLX 8-bit quantized conversion of Qwen/Qwen3-ASR-1.7B for Apple Silicon inference.

Usage

Used by qwen3-asr-swift Qwen3ASR module:

let model = try await Qwen3ASRModel.fromPretrained(
    modelId: "aufklarer/Qwen3-ASR-1.7B-MLX-8bit"
)
let text = model.transcribe(audio: samples, sampleRate: 16000)

audio transcribe --model large audio.wav

Model Details

• Architecture: Qwen3-ASR encoder-decoder (Whisper-style audio encoder + Qwen3 text decoder)
• Parameters: 1.7B
• Quantization: 8-bit (MLX)
• Size: ~2.3 GB
• Languages: Multilingual (EN, ZH, JA, KO, FR, DE, ES, and more)