Qwen3-ASR 0.6B (Apple Silicon)

Original model: Qwen/Qwen3-ASR-0.6B (Qwen / Alibaba)

On-device Qwen3-ASR (audio encoder + Qwen3 decoder) packaged by TheStage AI for Apple Silicon. Alternative ASR to Whisper for TheStage Apple SDK apps.


Runtime	TheStage Apple SDK (`Qwen3ASRPipeline`)
Input	16 kHz mono Float
HF engines	`TheStageAI/Qwen3-ASR-0.6B`

Overview

This repo hosts CoreML engine bundles for Qwen3-ASR. Primary shipping STT in the Apple SDK remains thewhisper-large-v3-turbo; use this model when you want the Qwen3-ASR family.

System Requirements

Property	Value
Hardware	Apple Silicon Mac, or physical iPhone / iPad
macOS	15.0+
iOS	18.0+
Xcode	16.0+
Swift	6.0+
Flutter (optional)	3.24+

Simulator is not supported — run on real Apple Silicon hardware.

TheStage AI Access Token

Create a token at app.thestage.ai and pass it to the SDK:

try await TheStageAI.shared.initialize(apiToken: "th_…")

await TheStageFlutterSDK.initialize(api_token: 'th_…');

Token is checked online in initialize (once per app process when reachable). Inference runs fully on-device. Offline initialize fails — reconnect and call initialize again.

TheStage Apple SDK

Installation (SwiftPM)

In Xcode: File → Add Package Dependencies…, paste https://github.com/TheStageAI/AppleSDK.git, and add the TheStageSDK product. Or in Package.swift:

.package(
    url: "https://github.com/TheStageAI/AppleSDK.git",
    exact: Version(1, 1, 0)
)

Installation (Flutter / iOS)

# pubspec.yaml
dependencies:
  thestage_apple_sdk:
    git:
      url: https://github.com/TheStageAI/AppleSDK.git
      path: plugin/thestage_apple_sdk
      ref: 1.1.0

Swift

import TheStageSDK

try await TheStageAI.shared.initialize(apiToken: "th_…")

let asr = try await Qwen3ASRPipeline(
    engines_path: "TheStageAI/Qwen3-ASR-0.6B"
)

// audio: 16 kHz mono Float in [-1, 1]
// language: "auto" (default) or a code / English name ("en", "English", …)
let result = asr.infer(audio: audio_samples, language: "auto")
print(result.text)

Flutter (iOS)

await TheStageFlutterSDK.start_model(
  model_name: 'stt',
  engines_path: 'TheStageAI/Qwen3-ASR-0.6B',
);

final result = await TheStageFlutterSDK.infer(
  model_name: 'stt',
  input_json: {
    'audio': audio_samples,  // Float32List, 16 kHz mono
    'language': 'auto',
  },
);
print(result[0]['transcription']);

Audio contract


Sample rate	16 kHz mono
Sample type	`Float` / `Float32List` in `[-1.0, 1.0]`

On-device latency

Release build on Apple M2 Max (NPU / ANE), macOS 26.2 (guidance, not an SLA):

Metric	Value
Audio fixture	8.88 s
RTFx	13.9
Decode tok/s	39.6
Process mem (MB)	~342

Acknowledgments

This work builds on Qwen3-ASR-0.6B by Qwen / Alibaba: Qwen/Qwen3-ASR-0.6B.

This Apple Silicon package is produced by TheStage AI; it is not an official Qwen release.

Model tree for TheStageAI/Qwen3-ASR-0.6B

Base model

Qwen/Qwen3-ASR-0.6B

Quantized

(38)

this model

Collection including TheStageAI/Qwen3-ASR-0.6B

iOS_Android_SDK

Collection

Bunch of models for deployment on iOS and macOS devices • 24 items • Updated 16 days ago

TheStageAI
/

Qwen3-ASR-0.6B