A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

Qwen3-ASR

Transcribes up to 30 seconds of 16 kHz mono speech locally, using sherpa-onnx INT8 on CPU or MLX Q4 on Apple Silicon GPU. CPU fallback is enabled by default and reports changes to device, precision and runtime.

Desktop inferencePrivate · Runs on your computer
About this model

Transcribes up to 30 seconds of 16 kHz mono speech locally, using sherpa-onnx INT8 on CPU or MLX Q4 on Apple Silicon GPU. CPU fallback is enabled by default and reports changes to device, precision and runtime.

License
Apache-2.0
Model released
Jan 29, 2026
Source repository created
Jan 28, 2026
Model weights · Q4
708.2 MB

Best for

  • Local transcription experiments
  • Comparing native and browser inference paths on the same recording

Keep in mind

  • The smoke test checks English only; other languages are not validated here.
  • Noise, accents, clipped speech, music and silence can cause omissions or hallucinations.
  • The prototype caps recordings at 30 seconds and generation at 256 tokens; fast speech can be truncated.
  • Review outputs; do not rely on them for safety-critical decisions.
  • Models remain loaded until clearMemory, terminate, deletion or app exit. Review transcripts; do not use them as authoritative safety-critical records.
  • The app downloads its own pinned CPython and isolated packages on request; installed Python, user packages and Python/pip configuration are not used. Automatic model-package provisioning is verified on Apple Silicon Mac; other platform stacks remain unverified.
  • Windows CUDA code is experimental, requires an explicitly installed pinned runtime and NVIDIA driver, and has not been tested on Windows hardware. DirectML/Vulkan/CoreML GPU implementations are not installed.
  • CPU fallback may change the requested precision to INT8. If the CPU runtime itself cannot initialize or infer, its error is returned.

Processed on your deviceYour input is processed locally in the desktop app. Model and runtime files download from external hosts.

Target platforms · verification required

  • macos · aarch64, x86_64
  • windows · x86_64

Recommended configuration

Q4 · gpu (verified platform only)

Sources & technical details
Original model
https://huggingface.co/Qwen/Qwen3-ASR-0.6B
Release verification
Verified
Release evidence
Official news explicitly releases the 0.6B and 1.7B ASR checkpoints.
Source updated
Jan 30, 2026
Native model files
https://github.com/k2-fsa/sherpa-onnx
Files updated
Apr 7, 2026
Artifact revision
sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96

Desktop app

Sample result

Qwen3 ASR 0.6B — Native Speech Recognition

Run Qwen3-ASR on your computer

Download the desktop app to load this model, provide input and view results locally.

Public desktop downloads are coming soon.

Device requirements

macOS Apple SiliconCPU / INT8Tested8.0 GiBRAM estimate
macOS Apple SiliconGPU / Q4Tested8.0 GiBRAM estimate
Windows Intel / AMDGPU / FP16UnverifiedNot establishedRAM estimate
linux Intel / AMDGPU / FP16UnverifiedNot establishedRAM estimate
Windows Intel / AMDCPU / INT8Unverified8.0 GiBRAM estimate
linux Intel / AMDCPU / INT8Unverified8.0 GiBRAM estimate
macOS Intel / AMDCPU / INT8Unverified8.0 GiBRAM estimate
If your device falls short

MemoryAdvisory only: close other apps, reduce image/tile/token/batch settings or choose a smaller verified model/precision. CPU can need more RAM than GPU.

GPUIf the selected GPU runtime fails, existing CPU fallback may recover when enabled, but CPU has its own memory/runtime needs.

PlatformUse a compatible OS/architecture/runtime or another model. These estimates do not add hardware detection or block runs.

Benchmark evidence

Saved desktop measurements apply to the recorded hardware and runtime. Fallback cases report the device that actually ran; they do not prove GPU support on the requested device.

Qwen3 ASR 0.6B — Native Speech Recognition

16 configurations tested
Passed16
Failed0
TestedNov 2, 2026
Runtimesherpa-onnx 2.0.1

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
Q4gpuAcceleratedPassedVery fastmacos · aarch6486 ms / 0 ms0.034 RTF
Q4gpuAcceleratedPassedVery fastmacos · aarch6487 ms / 0 ms0.034 RTF
Q4gpuAcceleratedPassedVery fastmacos · aarch6492 ms / 0 ms0.036 RTF
Q4gpuAcceleratedPassedVery fastmacos · aarch6481 ms / 0 ms0.044 RTF
Q4gpuAcceleratedPassedVery fastmacos · aarch64117 ms / 1385 ms0.046 RTF
Q4gpuAcceleratedPassedVery fastmacos · aarch64157 ms / 0 ms0.063 RTF
Q4gpuAcceleratedPassedFastmacos · aarch64679 ms / 1399 ms0.265 RTF
Q4gpuAcceleratedPassedFastmacos · aarch641290 ms / 1369 ms0.504 RTF
INT8cpuPassedFastmacos · aarch64544 ms / 0 ms0.212 RTF
INT8cpuPassedFastmacos · aarch64549 ms / 0 ms0.214 RTF
INT8cpuPassedFastmacos · aarch64558 ms / 1309 ms0.218 RTF
INT8cpuPassedFastmacos · aarch64559 ms / 1191 ms0.218 RTF
INT8cpuPassedFastmacos · aarch64570 ms / 0 ms0.223 RTF
INT8cpuPassedFastmacos · aarch64571 ms / 1307 ms0.223 RTF
INT8cpuPassedFastmacos · aarch64563 ms / 0 ms0.227 RTF
INT8cpuPassedFastmacos · aarch64495 ms / 0 ms0.269 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-02T11:04:03.016Z
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "runtimeProvisioning": "App-owned CPython3.13.18; model-owned pinned packages" }
Runtime
sherpa-onnx 2.0.1
Artifact revision
sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96
Fixture
{
  "file": "model-samples/desktop/restoration-variants/qwen-asr-17b-morning.wav",
  "url": "https://assets.localai.tools/samples/v1/files/f10adc7c63c21a15bbb5d015d243eabcb5ef5263bab238ae75cb002c36f69acd/qwen-asr-17b-morning.wav",
  "size": 81964,
  "sha1": "c5013bf9e74eede20a3ea7961d78f50ed056d2fc",
  "sha256": "f10adc7c63c21a15bbb5d015d243eabcb5ef5263bab238ae75cb002c36f69acd"
}
Rerun when
artifact checksum/revision; runtime version or provider build; input or quality gate; device/driver/OS/backend behavior; CPU thread selection or lifecycle change
Per-case quality & failure details
[
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 570.736875,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 570.736875,
      "loadMs": 1307.241958,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.22294409179687502,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 1877.9788330000001,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.485429472206435
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Please save the blue folder on your computer.",
      "transcript": "Please save the blue folder on your computer."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.48,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 563.078583,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 563.078583,
      "loadMs": 0,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.22704781572580646,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 563.078583,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.404358600866906
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Your code is four two seven.",
      "transcript": "Your code is four two seven."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 1.84,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 495.181209,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 495.181209,
      "loadMs": 0,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.26912022228260873,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 495.181209,
      "version": "2.0.1",
      "audioSecondsPerSecond": 3.715811437424718
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 570.2030420000001,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 570.2030420000001,
      "loadMs": 0,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.22273556328125005,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 570.2030420000001,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.48962880138405
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 558.4432499999999,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 558.4432499999999,
      "loadMs": 1309.2565000000004,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.21814189453124996,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 1867.6997500000002,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.584172160734328
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 543.880667,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 543.880667,
      "loadMs": 0,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.212453385546875,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 543.880667,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.706914871824264
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 559.248583,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 559.248583,
      "loadMs": 1191.2182500000008,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.21845647773437502,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 1750.4668330000009,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.577570829535745
    }
  },
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "cpu",
      "artifactRevision": "sha256:393f8a14e2f5fb96746aaab342997a40641001fbd5bf9592a080a8329178ee96",
      "audioSeconds": 2.56,
      "cpuThreads": 4,
      "device": "cpu",
      "durationMs": 549.0724580000001,
      "engine": "sherpa-onnx",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 549.0724580000001,
      "loadMs": 0,
      "modelId": "qwen3-asr-0.6b-native",
      "modelRetained": true,
      "precision": "int8",
      "provider": "CPUExecutionProvider",
      "realTimeFactor": 0.21448142890625005,
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "selectionReason": "Explicit verified CPU selection",
      "totalMs": 549.0724580000001,
      "version": "2.0.1",
      "audioSecondsPerSecond": 4.662408326443501
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 1290.2767909981776,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 1290.2767909981776,
      "loadMs": 1369.0761249999996,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1335226572,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.5040143714836631,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 2659.352915998177,
      "version": "0.6.3",
      "audioSecondsPerSecond": 1.9840704086597927
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Please save the blue folder on your computer.",
      "transcript": "Please save the blue folder on your computer."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.48,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 157.1561659948202,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 157.1561659948202,
      "loadMs": 0,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1420050524,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.06336942177210492,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 157.1561659948202,
      "version": "0.6.3",
      "audioSecondsPerSecond": 15.780481690306315
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Your code is four two seven.",
      "transcript": "Your code is four two seven."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 1.84,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 80.60012500209268,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 80.60012500209268,
      "loadMs": 0,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1406077468,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.043804415762006894,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 80.60012500209268,
      "version": "0.6.3",
      "audioSecondsPerSecond": 22.828748714127013
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 87.05299999564886,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 87.05299999564886,
      "loadMs": 0,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1416650852,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.034005078123300336,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 87.05299999564886,
      "version": "0.6.3",
      "audioSecondsPerSecond": 29.407372521658708
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 116.91745799907947,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 116.91745799907947,
      "loadMs": 1385.2244580000006,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1416517980,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.045670882030890425,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 1502.14191599908,
      "version": "0.6.3",
      "audioSecondsPerSecond": 21.895789078994135
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 86.31887499359436,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 86.31887499359436,
      "loadMs": 0,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1416644700,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.0337183105443728,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 86.31887499359436,
      "version": "0.6.3",
      "audioSecondsPerSecond": 29.65747642320379
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 678.672791007557,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 678.672791007557,
      "loadMs": 1398.6316669999997,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1380545228,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.26510655898732693,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 2077.3044580075566,
      "version": "0.6.3",
      "audioSecondsPerSecond": 3.7720681216634993
    }
  },
  {
    "precision": "q4",
    "device": "gpu",
    "status": "pass",
    "quality": {
      "wordAccuracy": 1,
      "wordErrorRate": 0,
      "repetitive": false,
      "passed": true,
      "expected": "Good morning. The train leaves at nine.",
      "transcript": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "acceleration": "partial",
      "accelerationDetails": "MLX model computation on Metal; CPU preprocessing/tokenization",
      "artifactRevision": "313d850181767edf09f00a9c289becca70e58cd0",
      "audioSeconds": 2.56,
      "cpuThreads": null,
      "cpuThreadsEffect": "CPU thread setting applies to the sherpa CPU backend, not the MLX/CUDA GPU runtime",
      "device": "gpu",
      "durationMs": 92.46804199938197,
      "engine": "mlx-audio",
      "fallback": false,
      "fallbackReason": null,
      "grade": "good",
      "inferenceMs": 92.46804199938197,
      "loadMs": 0,
      "mlxVersion": "0.32.4",
      "modelId": "qwen3-asr-0.6b-native",
      "peakMemoryBytes": 1416632924,
      "precision": "q4",
      "provider": "Metal",
      "realTimeFactor": 0.03612032890600858,
      "requestedCpuThreads": 4,
      "requestedDevice": "gpu",
      "requestedPrecision": "q4",
      "selectedDevice": "Device(gpu, 0)",
      "totalMs": 92.46804199938197,
      "version": "0.6.3",
      "audioSecondsPerSecond": 27.685240702048286
    }
  }
]

Qwen3-ASR1.7B · Native speech recognition

33 configurations tested
Passed33
Failed0
TestedNov 1, 2026
Runtimetransformers 5.22.1

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedFastmacos · aarch64475 ms / 0 ms0.186 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64478 ms / 0 ms0.187 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64480 ms / 0 ms0.187 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64480 ms / 0 ms0.187 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64480 ms / 0 ms0.187 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64494 ms / 0 ms0.193 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64660 ms / 1212 ms0.258 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64683 ms / 1236 ms0.267 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64688 ms / 0 ms0.277 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64640 ms / 0 ms0.348 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch641388 ms / 1271 ms0.542 RTF
FP32cpuPassedFastmacos · aarch641141 ms / 0 ms0.446 RTF
FP32cpuPassedFastmacos · aarch641145 ms / 0 ms0.447 RTF
FP32cpuPassedFastmacos · aarch641154 ms / 0 ms0.451 RTF
FP32cpuPassedFastmacos · aarch641162 ms / 0 ms0.454 RTF
FP32cpuPassedFastmacos · aarch641164 ms / 0 ms0.455 RTF
FP32cpuPassedFastmacos · aarch641167 ms / 0 ms0.456 RTF
FP32cpuPassedFastmacos · aarch641175 ms / 647 ms0.459 RTF
FP32cpuPassedFastmacos · aarch641140 ms / 0 ms0.460 RTF
FP32cpuPassedFastmacos · aarch641178 ms / 679 ms0.460 RTF
FP32cpuPassedFastmacos · aarch641189 ms / 0 ms0.465 RTF
FP32cpuPassedFastmacos · aarch641202 ms / 0 ms0.469 RTF
FP32cpuPassedFastmacos · aarch641222 ms / 0 ms0.478 RTF
FP32cpuPassedFastmacos · aarch641208 ms / 0 ms0.487 RTF
FP32cpuPassedFastmacos · aarch641250 ms / 615 ms0.488 RTF
FP32cpuPassedFastmacos · aarch641250 ms / 607 ms0.488 RTF
FP32cpuPassedFastmacos · aarch641253 ms / 0 ms0.490 RTF
FP32cpuPassedFastmacos · aarch641254 ms / 682 ms0.490 RTF
FP32cpuPassedFastmacos · aarch641254 ms / 594 ms0.490 RTF
FP32cpuPassedFastmacos · aarch641261 ms / 0 ms0.493 RTF
FP32cpuPassedFastmacos · aarch641270 ms / 0 ms0.496 RTF
FP32cpuPassedModeratemacos · aarch64992 ms / 0 ms0.539 RTF
FP32cpuPassedModeratemacos · aarch641074 ms / 0 ms0.584 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-01T23:20:11.749396+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2" }
Runtime
transformers 5.22.1
Artifact revision
bcd2b5b7f32b480ab5790554cfa8347f246a14f3
Fixture
{
  "file": "model-samples/desktop/restoration-variants/qwen-asr-17b-morning.wav",
  "url": "https://assets.localai.tools/samples/v1/files/f10adc7c63c21a15bbb5d015d243eabcb5ef5263bab238ae75cb002c36f69acd/qwen-asr-17b-morning.wav",
  "size": 81964,
  "sha1": "c5013bf9e74eede20a3ea7961d78f50ed056d2fc",
  "sha256": "f10adc7c63c21a15bbb5d015d243eabcb5ef5263bab238ae75cb002c36f69acd"
}
Rerun when
artifact checksum/revision; runtime version or provider build; input or quality gate; device/driver/OS/backend behavior; CPU thread selection or lifecycle change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 678.973957998096,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1188.7231660002726,
      "inferenceMs": 1177.5432919966988,
      "realTimeFactor": 0.4599778484362105,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1867.6971239983686,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Please save the blue folder on your computer."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1141.4814579984522,
      "inferenceMs": 1140.1885000013863,
      "realTimeFactor": 0.4597534274199138,
      "audioSeconds": 2.48,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1141.4814579984522,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Your code is four two seven."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 993.6614589969395,
      "inferenceMs": 992.3387079979875,
      "realTimeFactor": 0.5393145152162976,
      "audioSeconds": 1.84,
      "outputPlacement": "cpu",
      "outputTokens": 11,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 993.6614589969395,
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1262.3185420015943,
      "inferenceMs": 1261.0690830042586,
      "realTimeFactor": 0.4926051105485385,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1262.3185420015943,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1223.7862499969197,
      "inferenceMs": 1222.4378749961033,
      "realTimeFactor": 0.47751479492035287,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1223.7862499969197,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 647.1819170037634,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1185.2696670030127,
      "inferenceMs": 1175.3664999996545,
      "realTimeFactor": 0.45912753906236503,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1832.451584006776,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1154.9236670034588,
      "inferenceMs": 1153.651416003413,
      "realTimeFactor": 0.4506450843763332,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1154.9236670034588,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1142.3365000009653,
      "inferenceMs": 1140.8314580039587,
      "realTimeFactor": 0.44563728828279636,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1142.3365000009653,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 681.5112499971292,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1263.6313749972032,
      "inferenceMs": 1253.8412080029957,
      "realTimeFactor": 0.4897817218761702,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1945.1426249943324,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1168.7504579967936,
      "inferenceMs": 1167.4625410014414,
      "realTimeFactor": 0.45604005507868806,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1168.7504579967936,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1146.63704099803,
      "inferenceMs": 1145.2765420035576,
      "realTimeFactor": 0.44737364922013967,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1146.63704099803,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 607.1135000020149,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1261.081083001045,
      "inferenceMs": 1250.3776669982472,
      "realTimeFactor": 0.4884287761711903,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1868.19458300306,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Please save the blue folder on your computer."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1209.9585410032887,
      "inferenceMs": 1208.3954580011778,
      "realTimeFactor": 0.48725623306499105,
      "audioSeconds": 2.48,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1209.9585410032887,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Your code is four two seven."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1075.4037080041599,
      "inferenceMs": 1074.1128329973435,
      "realTimeFactor": 0.583756974455078,
      "audioSeconds": 1.84,
      "outputPlacement": "cpu",
      "outputTokens": 11,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1075.4037080041599,
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1254.8535000023548,
      "inferenceMs": 1253.1545419988106,
      "realTimeFactor": 0.4895134929682854,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1254.8535000023548,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1202.968666999368,
      "inferenceMs": 1201.54754199757,
      "realTimeFactor": 0.4693545085928008,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1202.968666999368,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 594.2322080009035,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1264.2030829956639,
      "inferenceMs": 1254.2732909932965,
      "realTimeFactor": 0.48995050429425646,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1858.4352909965673,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1163.8499580003554,
      "inferenceMs": 1161.9883329985896,
      "realTimeFactor": 0.45390169257757407,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1163.8499580003554,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1165.2754999959143,
      "inferenceMs": 1163.919958002225,
      "realTimeFactor": 0.45465623359461915,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1165.2754999959143,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 614.969750000455,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1260.4346250009257,
      "inferenceMs": 1250.228667006013,
      "realTimeFactor": 0.4883705730492238,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1875.4043750013807,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1191.0225839965278,
      "inferenceMs": 1189.3087079952238,
      "realTimeFactor": 0.4645737140606343,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1191.0225839965278,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1271.4551250028308,
      "inferenceMs": 1269.9637909972807,
      "realTimeFactor": 0.4960796058583128,
      "audioSeconds": 2.56,
      "outputPlacement": "cpu",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1271.4551250028308,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 1271.2682909987052,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 1399.57320800022,
      "inferenceMs": 1388.2919999959995,
      "realTimeFactor": 0.5423015624984373,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 2670.841498998925,
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Please save the blue folder on your computer."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 692.0158339999034,
      "inferenceMs": 687.9505420001806,
      "realTimeFactor": 0.277399412096847,
      "audioSeconds": 2.48,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8938045440,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 692.0158339999034,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Your code is four two seven."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 643.4208749997197,
      "inferenceMs": 640.1404579955852,
      "realTimeFactor": 0.3479024228236876,
      "audioSeconds": 1.84,
      "outputPlacement": "mps:0",
      "outputTokens": 11,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8939847680,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 643.4208749997197,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 497.49891700048465,
      "inferenceMs": 493.9422919997014,
      "realTimeFactor": 0.19294620781238336,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8939847680,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 497.49891700048465,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 483.594292003545,
      "inferenceMs": 479.88512499432545,
      "realTimeFactor": 0.18745512695090838,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8939847680,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 483.594292003545,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 1211.8349169977591,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 672.0745829952648,
      "inferenceMs": 659.7911659991951,
      "realTimeFactor": 0.2577309242184356,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1883.909499993024,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 482.98433300078614,
      "inferenceMs": 479.6113749980577,
      "realTimeFactor": 0.1873481933586163,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 482.98433300078614,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 483.63187500217464,
      "inferenceMs": 479.62616599397734,
      "realTimeFactor": 0.1873539710913974,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 483.63187500217464,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 1235.509792000812,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 694.6250420005526,
      "inferenceMs": 683.2100410028943,
      "realTimeFactor": 0.2668789222667556,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 1930.1348340013647,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 478.9757920007105,
      "inferenceMs": 475.4872909979895,
      "realTimeFactor": 0.18573722304608964,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 478.9757920007105,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.25,
      "repetitive": false,
      "expected": "Good morning. The train leaves at nine."
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 707
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Audio encoder and language decoder parameters/input/output IDs on selected device; audio features, tokenization and transcript parsing on CPU",
      "durationMs": 481.6857080004411,
      "inferenceMs": 477.82570899289567,
      "realTimeFactor": 0.18665066757534987,
      "audioSeconds": 2.56,
      "outputPlacement": "mps:0",
      "outputTokens": 13,
      "decoding": "greedy",
      "sampleRate": 16000,
      "driverAllocatedMemoryBytes": 8935931904,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "bcd2b5b7f32b480ab5790554cfa8347f246a14f3",
      "totalMs": 481.6857080004411,
      "fallback": false,
      "grade": "fast"
    }
  }
]

Something not quite right?

Tell us what happened with Qwen3-ASR. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.