Modelpublic

appautomaton/qwen3-asr-1.7b-bf16-mlx

sourceHugging Faceapache-2.0updated 1h agoView on Hugging Face

0likes32downloads

Model Card

Qwen3-ASR-1.7B — MLX (bf16)

![GitHub](https://github.com/appautomaton/mlx-speech) ![App Automaton](https://appautomaton.com) ![Hugging Face](https://huggingface.co/appautomaton) ![int8 build-1f6feb)](https://huggingface.co/appautomaton/qwen3-asr-1.7b-int8-mlx)

MLX-native bf16 conversion of Qwen3-ASR-1.7B for local automatic speech recognition on Apple Silicon — English, Chinese, and mixed Chinese/English. It runs through the `mlx-speech` runtime with no PyTorch and no cloud API at inference time. Weights ship as plain .safetensors.

This is the unquantized reference build. Most users should prefer the **int8 build** — it matched this build's transcripts on our English test clips. Use bf16 when you want the full-precision baseline (e.g. to compare quantization quality yourself).

Model Details

—Developed by: App Automaton
—Upstream model: `Qwen/Qwen3-ASR-1.7B` (code: `QwenLM/Qwen3-ASR`)
—Task: automatic speech recognition — English, Chinese, and mixed Chinese/English (upstream supports more languages; the mlx-speech v0 path validates en/zh/mixed)
—Precision: bf16 — MLX format port; keys are remapped to the MLX module tree and audio Conv2D weights transposed to MLX layout
—Runtime: MLX on Apple Silicon
—Input: 16 kHz mono audio

Variants

Repo	Precision	Notes
`qwen3-asr-1.7b-bf16-mlx` (this repo)	bf16	unquantized reference
`qwen3-asr-1.7b-int8-mlx`	int8 affine	default — recommended for most users

How to Get Started

Install `mlx-speech`, then load by alias (downloads on first use):

python

import mlx_speech

asr = mlx_speech.asr.load("qwen3-asr-1.7b-bf16")
result = asr.generate("speech.wav")               # language=None auto-detects
print(result.language, result.text)

# For mixed Chinese/English, forcing Chinese preserves Chinese characters best:
result = asr.generate("mixed-speech.wav", language="Chinese")

bash

mlx-speech asr --model qwen3-asr-1.7b-bf16 --audio speech.wav

Or download the weights once and load by local path:

bash

hf download appautomaton/qwen3-asr-1.7b-bf16-mlx \
  --local-dir models/qwen3_asr_1_7b/mlx-bf16

python

asr = mlx_speech.asr.load("models/qwen3_asr_1_7b/mlx-bf16")

Notes

—This repo contains the MLX runtime artifact only (no PyTorch checkpoint).
—v0 is single-pass offline ASR; streaming, timestamps, and long-audio chunking are deferred.
—For a smaller download, use the int8 build.

License