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.

Apertus 1.5

Generate text from a conversation, a still image, a WAV clip, or a combination of those inputs.

Desktop inferencePrivate · Runs on your computer
About this model

Generate text from a conversation, a still image, a WAV clip, or a combination of those inputs.

License
Apache-2.0
Model released
Unknown
Source repository created
Jul 24, 2026
Model weights · BF16
18.4 GB

Best for

  • Local multilingual chat, image questions and short speech transcription.

Keep in mind

  • Outputs can be inaccurate, fabricated, biased or unsafe. Review answers before relying on them.
  • Poor lighting, low resolution and noisy audio can reduce accuracy. Still PNG/JPEG and short WAV clips only.
  • The adapter limits context to 8192 tokens and uses greedy decoding with thinking disabled.
  • The official Hugging Face repository requires access for the account supplying the download credential.
  • Local results cover the measured Apple Silicon Mac. Windows and Intel Mac are untested.

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

BF16 · mps (verified platform only)

Sources & technical details
Original model
https://huggingface.co/swiss-ai/Apertus-v1.5-8B
Release verification
Unknown
Release evidence
Exact checkpoint publication day not established. Hub creation is retained separately.
Source updated
Nov 3, 2026
Native model files
https://huggingface.co/swiss-ai/Apertus-v1.5-8B
Files updated
Nov 3, 2026
Artifact revision
1519de0ae7f3d114879291810a58259966f64c99

Desktop app

Sample result

Apertus v1.5 8B

Run Apertus 1.5 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 SiliconMPS / BF16Tested32.0 GiBRAM estimate
macOS Apple SiliconCPU / FP32Tested48.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested48.0 GiBRAM estimate
macOS Apple SiliconCPU / BF16Unverified32.0 GiBRAM estimate
Windows Intel / AMDCPU / FP32Unverified48.0 GiBRAM estimate
macOS Intel / AMDCPU / FP32Unverified48.0 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.

Apertus v1.5 8B

45 configurations tested
Passed44
Failed1
TestedNov 13, 2026
Runtimetransformers 5.14.0.dev0

Quality criteria

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

PrecisionDeviceResultPlatformInference / load
BF16mpsAcceleratedPassedFastmacos · aarch64354 ms / 5765 ms
BF16mpsAcceleratedPassedFastmacos · aarch64380 ms / 6410 ms
BF16mpsAcceleratedPassedFastmacos · aarch641782 ms / 6410 ms
BF16mpsAcceleratedPassedFastmacos · aarch642853 ms / 6410 ms
BF16mpsAcceleratedPassedFastmacos · aarch64487 ms / 5765 ms
FP32mpsAcceleratedPassedFastmacos · aarch64542 ms / 10656 ms
FP32mpsAcceleratedPassedFastmacos · aarch64547 ms / 11705 ms
BF16mpsAcceleratedPassedFastmacos · aarch64681 ms / 5765 ms
FP32mpsAcceleratedPassedFastmacos · aarch64834 ms / 10656 ms
BF16mpsAcceleratedPassedFastmacos · aarch641144 ms / 6410 ms
BF16mpsAcceleratedPassedFastmacos · aarch641717 ms / 6410 ms
BF16mpsAcceleratedPassedFastmacos · aarch642117 ms / 6410 ms
FP32mpsAcceleratedPassedFastmacos · aarch642483 ms / 11705 ms
FP32mpsAcceleratedPassedSlowmacos · aarch643178 ms / 11705 ms
FP32mpsAcceleratedPassedSlowmacos · aarch647944 ms / 11705 ms
FP32mpsAcceleratedPassedSlowmacos · aarch643169 ms / 11705 ms
FP32mpsAcceleratedPassedSlowmacos · aarch645046 ms / 11705 ms
FP32mpsAcceleratedPassedSlowmacos · aarch647491 ms / 10656 ms
FP32cpuPassedFastmacos · aarch641498 ms / 8268 ms
FP32cpuPassedFastmacos · aarch641536 ms / 4193 ms
FP32cpuPassedFastmacos · aarch641564 ms / 5685 ms
FP32cpuPassedFastmacos · aarch641673 ms / 8268 ms
FP32cpuPassedFastmacos · aarch641689 ms / 5685 ms
FP32cpuPassedFastmacos · aarch642066 ms / 4193 ms
FP32cpuPassedFastmacos · aarch642212 ms / 5685 ms
FP32cpuPassedFastmacos · aarch642252 ms / 8268 ms
FP32cpuPassedFastmacos · aarch642881 ms / 8732 ms
FP32cpuPassedSlowmacos · aarch644708 ms / 4193 ms
FP32cpuPassedSlowmacos · aarch644795 ms / 8732 ms
FP32cpuPassedSlowmacos · aarch648749 ms / 4193 ms
FP32cpuPassedSlowmacos · aarch649532 ms / 8732 ms
FP32cpuPassedSlowmacos · aarch644260 ms / 4193 ms
FP32cpuPassedSlowmacos · aarch644572 ms / 4193 ms
FP32cpuPassedSlowmacos · aarch644783 ms / 8732 ms
FP32cpuPassedSlowmacos · aarch646914 ms / 8732 ms
FP32cpuPassedSlowmacos · aarch647212 ms / 8732 ms
BF16cpuPassedSlowmacos · aarch648527 ms / 3097 ms
BF16cpuPassedSlowmacos · aarch649139 ms / 3097 ms
BF16cpuPassedSlowmacos · aarch649254 ms / 3332 ms
BF16cpuPassedSlowmacos · aarch649299 ms / 3097 ms
BF16cpuPassedSlowmacos · aarch6412933 ms / 3332 ms
BF16cpuPassedVery slowmacos · aarch6423614 ms / 3332 ms
BF16cpuPassedVery slowmacos · aarch6440915 ms / 3332 ms
BF16cpuPassedVery slowmacos · aarch6444035 ms / 3332 ms
BF16cpuFailed—macos · aarch6458074 ms / 3332 ms
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-13T17:03:50.494262+00:00
Environment
{ "os": "macos", "architecture": "aarch64", "osVersion": "27.2", "hardware": "Apple M3 Max; 96 GiB", "runtime": "Swiss AI Transformers fork 3797303dda74844e3d1f8977ff5518bb91f818b4 with Torch 2.10.0", "concurrentUploadActivity": "none", "timingScope": "Local pinned snapshot, strict device placement, two helper starts and repeated text requests. Excludes transfer." }
Runtime
transformers 5.14.0.dev0
Artifact revision
1519de0ae7f3d114879291810a58259966f64c99
Fixture
{
  "file": "research/apertus-v1.5/matrix-fixtures.json",
  "size": 2212,
  "sha1": "412aba9defe7423dfc7b431e4af78c9b55f98eb2",
  "sha256": "c86054aaffe62bd4eb4a72acb6097d1405dd7c0e1c7c9ad6c7ed8aa38af73360",
  "url": "https://assets.localai.tools/samples/v1/files/c86054aaffe62bd4eb4a72acb6097d1405dd7c0e1c7c9ad6c7ed8aa38af73360/matrix-fixtures.json"
}
Rerun when
artifact revision; runtime; fixture or quality gate; device/provider behavior
Per-case quality & failure details
[
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1717.4195420229807,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 1.1645378144726142,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 8127.576584054623,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "berlin",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 380.4316249443218,
      "inputTokens": 79,
      "outputTokens": 2,
      "tokensPerSecond": 5.257186492560156,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 6790.539667010307,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "red",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2116.5768330101855,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.944921993290274,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 8526.747417054139,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1144.2707909736782,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 1.7478380255587649,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 7554.417125007603,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "hello this is a local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1781.9801659788936,
      "inputTokens": 199,
      "outputTokens": 9,
      "tokensPerSecond": 5.050561264275373,
      "streamChunks": 8,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 8201.901166990865,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue|local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 6409.971625020262,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2852.6648330152966,
      "inputTokens": 476,
      "outputTokens": 13,
      "tokensPerSecond": 4.557142447841958,
      "streamChunks": 9,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 9272.111250029411,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5765.454499982297,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 681.090584024787,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 2.93646696476311,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 6446.65320898639,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5765.454499982297,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 353.9687079610303,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 5.650216968388594,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 6119.550083996728,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5765.454499982297,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.bfloat16": 387,
        "mps:0/torch.float32": 469
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 487.35124996164814,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 4.103816292986607,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "bf16",
      "totalMs": 6252.925124950707,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 5045.953374996316,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.396357209702014,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 16750.71612506872,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "berlin",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 547.2708329907618,
      "inputTokens": 79,
      "outputTokens": 2,
      "tokensPerSecond": 3.654497699192679,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 12251.987042021938,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "red",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 3169.0501249977387,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.6311039337067688,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 14874.558501003776,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2483.040541992523,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.8054640937900653,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 14188.097917067353,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "hello this is a local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 3177.648124983534,
      "inputTokens": 199,
      "outputTokens": 9,
      "tokensPerSecond": 2.832283388849618,
      "streamChunks": 8,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 14899.10120802233,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue|local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 11704.55916703213,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 7944.497124990448,
      "inputTokens": 476,
      "outputTokens": 13,
      "tokensPerSecond": 1.6363527855157516,
      "streamChunks": 9,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 19667.30783402454,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 10656.292417028453,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 7490.852291986812,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.26699231569943777,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 18147.33479201095,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 10656.292417028453,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 542.3344159498811,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 3.6877615382329463,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 11198.847333027516,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 10656.292417028453,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "mps:0/torch.float32": 856
      },
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 833.5129170445725,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 2.399482910344686,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "mps",
      "precision": "fp32",
      "totalMs": 11490.054167050403,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 12933.417166001163,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.15463817290742912,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 16265.57191601023,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "berlin",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 9254.059374972712,
      "inputTokens": 79,
      "outputTokens": 2,
      "tokensPerSecond": 0.2161213710611077,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 12586.102499975823,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "red",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 40915.1316669886,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.048881670876148066,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 44247.26049997844,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "very-slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 44035.339209018275,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.045418067305143126,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 47367.44208401069,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "very-slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "hello this is a local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 23613.92249999335,
      "inputTokens": 199,
      "outputTokens": 9,
      "tokensPerSecond": 0.38113108908537047,
      "streamChunks": 8,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 26955.18420904409,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "very-slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "fail",
    "quality": {
      "passed": false,
      "expected": "blue|local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3331.9366250070743,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 58074.218875030056,
      "inputTokens": 476,
      "outputTokens": 2,
      "tokensPerSecond": 0.03443868964133295,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 61416.72308300622,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "very-slow"
    },
    "failure": "Measured response missed the frozen quality or latency gate."
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3096.7552909860387,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 9138.86058301432,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.2188456626329591,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 12235.747749975417,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3096.7552909860387,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 8527.41237502778,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.23453773689389387,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 11624.288624967448,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "bf16",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 3096.7552909860387,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.bfloat16": 387,
        "cpu/torch.float32": 469
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 9298.566209035926,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.21508692362232046,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "bf16",
      "totalMs": 12395.424832007848,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 6913.686166983098,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.2892812823282566,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 15646.809292025864,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "berlin",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2880.854958028067,
      "inputTokens": 79,
      "outputTokens": 2,
      "tokensPerSecond": 0.6942383525510745,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 11613.344998972025,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "red",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 7211.632416001521,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.2773297201840602,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 15944.099624990486,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 4782.895499956794,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.41815674208605785,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 13515.1766660274,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "hello this is a local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 4794.633374956902,
      "inputTokens": 199,
      "outputTokens": 9,
      "tokensPerSecond": 1.8770986843349413,
      "streamChunks": 8,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 13536.026999005117,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue|local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8732.128583011217,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 9531.523917044979,
      "inputTokens": 476,
      "outputTokens": 13,
      "tokensPerSecond": 1.3638952294661335,
      "streamChunks": 9,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 18273.412291018758,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8267.974207992665,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2251.9980000215583,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.888100255853182,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 10520.104041032027,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8267.974207992665,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1673.3389170258306,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 1.195215135230807,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 9941.43012497807,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 8267.974207992665,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1498.4953330131248,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 1.3346721580897194,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 9766.554083034862,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2066.307708970271,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.9679100510139823,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 6259.856541000772,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "berlin",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1536.095374962315,
      "inputTokens": 79,
      "outputTokens": 2,
      "tokensPerSecond": 1.3020024879959462,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 5729.602415929548,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "red",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 4571.601249976084,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.43748347474759813,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 8765.209582925308,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 4260.462209000252,
      "inputTokens": 353,
      "outputTokens": 2,
      "tokensPerSecond": 0.46943263474441527,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 8454.025082930457,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "hello this is a local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 4708.373915986158,
      "inputTokens": 199,
      "outputTokens": 9,
      "tokensPerSecond": 1.9114879490438623,
      "streamChunks": 8,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 8911.347832996398,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "blue|local speech synthesis test",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 4193.411415966693,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 8749.051916995086,
      "inputTokens": 476,
      "outputTokens": 13,
      "tokensPerSecond": 1.4858752837833116,
      "streamChunks": 9,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 12951.863832015079,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5684.639166982379,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 2212.3847500188276,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 0.904001892068267,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 7897.122125024907,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5684.639166982379,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1564.0456249821,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 1.278735075278184,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 7248.780875990633,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "expected": "paris",
      "method": "Frozen normalized answer and stream parity gate"
    },
    "runtime": {
      "loadMs": 5684.639166982379,
      "version": "5.14.0.dev0",
      "torchVersion": "2.10.0",
      "placement": {
        "cpu/torch.float32": 856
      },
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Neural execution on the selected device. Media decoding, resampling and tokenization use CPU.",
      "inferenceMs": 1688.9575840323232,
      "inputTokens": 76,
      "outputTokens": 2,
      "tokensPerSecond": 1.1841623608007221,
      "streamChunks": 1,
      "engine": "transformers",
      "device": "cpu",
      "precision": "fp32",
      "totalMs": 7373.700249940157,
      "artifactRevision": "1519de0ae7f3d114879291810a58259966f64c99",
      "fallback": false,
      "grade": "fast"
    }
  }
]
Release validation still required
  • Uploads excluded. Public sample assets, full backup and forced source recovery pending.

Something not quite right?

Tell us what happened with Apertus 1.5. 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.