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.

Granite Docling Desktop

Convert page images or locally rendered PDFs into DocTags, Markdown and structured DoclingDocument JSON using native Granite Docling and Docling Core.

Desktop inferencePrivate · Runs on your computer
About this model

Convert page images or locally rendered PDFs into DocTags, Markdown and structured DoclingDocument JSON using native Granite Docling and Docling Core.

Created by
IBM Granite
License
Apache-2.0
Model released
Sep 17, 2025
Source repository created
May 19, 2025
Model weights · FP32
515.1 MB

Best for

  • Local document text/layout/table extraction with review
  • Converting scanned pages/PDFs to structured Docling JSON and Markdown
  • Document conversion experiments and annotated benchmarks

Keep in mind

  • OCR, reading order, table cells/structure, coordinates and formulas can be wrong, omitted, duplicated or hallucinated; review against original documents.
  • Not legal/medical/financial/identity/accessibility-critical ground truth, and not guaranteed exact transcription or complete privacy redaction.
  • English-oriented model; other languages, handwriting, poor scans, small text and complex layouts are not certified by this small authored suite.
  • Disabling image splitting failed receipt content/completion on CPU and Metal; doImageSplitting=false is rejected before inference.
  • PDFs are rasterized locally with PDFium then read page by page; embedded text is not a separate authoritative extraction path. Encrypted/malformed/truncated files fail validation.
  • A per-page maxNewTokens budget can truncate generation. Runtime records each page EOS/limit status; partial DocTags/Markdown/JSON are not guaranteed full-page extraction.
  • Accepts still PNG/JPEG and PDFs. No product page-count/duration cap; memory/time grow with image resolution, PDF render scale, patch splitting, document length and output.
  • Structured output omits duplicate page image bytes while retaining page size, layout geometry and provenance. DocTags/model-produced content is untrusted and never executed.
  • Default path is the official native Transformers Granite VLM plus Docling Core conversion, not every optional Docling standard-pipeline/OCR/plugin feature.
  • Only FP32 CPU/Metal advertised on measured Mac. Metal FP16 failed content/completion and BF16 failed table cells/structure; both are unsupported and preserved as failed candidates. Windows/Intel/other Apple versions and vendors remain unverified.
  • 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.
  • Completeness limitation: disabling doImageSplitting with maxNewTokens=1024 truncated the receipt on CPU and Metal without EOS. Use the tested splitting-enabled profile for that fixture; inspect output-limit/EOS metadata and never treat a valid document shape as complete extraction.

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

FP32 · mps (verified platform only)

Sources & technical details
Original model
https://huggingface.co/ibm-granite/granite-docling-258M
Release verification
Verified
Release evidence
Exact checkpoint Release Date stated by original provider in model card.
Source updated
Sep 23, 2025
Native model files
https://huggingface.co/ibm-granite/granite-docling-258M
Files updated
Sep 23, 2025
Artifact revision
982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe

Desktop app

Sample result

Docling — Image to Text Desktop

Run Granite Docling Desktop 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 / FP32Tested24.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested24.0 GiBRAM estimate
macOS Apple SiliconMPS / BF16Unverified24.0 GiBRAM estimate
Windows Intel / AMDCPU / FP32Unverified24.0 GiBRAM estimate
macOS Intel / AMDCPU / FP32Unverified24.0 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.

Docling — Image to Text Desktop

37 configurations tested
Passed31
Failed6
TestedNov 1, 2026
Runtimetransformers 5.22.1

Quality criteria

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

PrecisionDeviceResultPlatformInference / load
FP32mpsAcceleratedPassedFastmacos · aarch641120 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch641938 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642262 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642617 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642639 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642659 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642680 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642744 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642757 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642768 ms / 0 ms
FP32mpsAcceleratedPassedFastmacos · aarch642836 ms / 400 ms
FP32mpsAcceleratedPassedFastmacos · aarch642873 ms / 463 ms
FP32mpsAcceleratedPassedFastmacos · aarch642919 ms / 452 ms
FP32mpsAcceleratedPassedFastmacos · aarch644646 ms / 0 ms
BF16mpsAcceleratedPassedFastmacos · aarch644728 ms / 642 ms
FP32cpuPassedFastmacos · aarch642996 ms / 0 ms
FP32cpuPassedFastmacos · aarch644297 ms / 0 ms
FP32cpuPassedFastmacos · aarch645075 ms / 0 ms
FP32cpuPassedFastmacos · aarch645859 ms / 0 ms
FP32cpuPassedFastmacos · aarch645925 ms / 0 ms
FP32cpuPassedFastmacos · aarch645929 ms / 0 ms
FP32cpuPassedFastmacos · aarch645951 ms / 0 ms
FP32cpuPassedFastmacos · aarch645990 ms / 0 ms
FP32cpuPassedFastmacos · aarch646042 ms / 0 ms
FP32cpuPassedFastmacos · aarch646048 ms / 0 ms
FP32cpuPassedFastmacos · aarch646049 ms / 233 ms
FP32cpuPassedFastmacos · aarch646050 ms / 231 ms
FP32cpuPassedFastmacos · aarch646079 ms / 234 ms
FP32cpuPassedFastmacos · aarch649914 ms / 0 ms
FP32cpuPassedModeratemacos · aarch6410072 ms / 222 ms
FP32cpuPassedModeratemacos · aarch6411126 ms / 0 ms
FP16mpsFailed—macos · aarch6413651 ms / 399 ms
FP16mpsFailed—macos · aarch6413649 ms / 0 ms
FP16mpsFailed—macos · aarch6430191 ms / 0 ms
BF16mpsFailed—macos · aarch643414 ms / 0 ms
FP32cpuFailed—macos · aarch6417932 ms / 229 ms
FP32mpsFailed—macos · aarch6411112 ms / 402 ms
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-01T18:24:15.587596+00:00
Environment
{ "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "python": "3.13.1", "runtime": "Transformers5.22.1/Torch2.10.0/DoclingCore2.105.0/PDFium5.14.0" }
Runtime
transformers 5.22.1
Artifact revision
982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe
Fixture
{
  "file": "model-samples/desktop/docling/receipt.png",
  "size": 25773,
  "sha1": "05f77f558dde3e17f3541910abeecde08c193cd5",
  "sha256": "cc0123c5e9e48de091ce0d3462d267b2485295096206353c9bf4beeec594ca0e",
  "url": "https://assets.localai.tools/samples/v1/files/cc0123c5e9e48de091ce0d3462d267b2485295096206353c9bf4beeec594ca0e/receipt.png"
}
Rerun when
artifact revision/checksums; Transformers/Torch/DoclingCore/PDFium runtime or driver/backend changes; PDF rendering/image processor/DocTags conversion/output budget changes; fixtures/independent content/table annotations/quality gates
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 221.53637500014156,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 10546.207333001803,
      "inferenceMs": 10071.50191600158,
      "pagesPerSecond": 0.09929005706797336,
      "tokensPerSecond": 15.687829016739792,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 10071.50191600158,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 10767.743708001944,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 10054.900209001062,
      "inferenceMs": 9914.482208998379,
      "pagesPerSecond": 0.10086255428371242,
      "tokensPerSecond": 15.936283576826561,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 9914.482208998379,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 10054.900209001062,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 232.50833400015836,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6573.0403330017,
      "inferenceMs": 6049.472582999442,
      "pagesPerSecond": 0.1653036667709272,
      "tokensPerSecond": 26.117979349806497,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6049.472582999442,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6805.548667001858,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 1,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 5221.63804199954,
      "inferenceMs": 5074.659333000454,
      "pagesPerSecond": 0.19705756276032382,
      "tokensPerSecond": 22.85867728019756,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 116,
          "inferenceMs": 5074.659333000454,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 116,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 5221.63804199954,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 11417.920667001454,
      "inferenceMs": 11125.83399900177,
      "pagesPerSecond": 0.17976180483902993,
      "tokensPerSecond": 22.739868312137286,
      "pageCount": 2,
      "pages": [
        {
          "page": 1,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6258.920791002311,
          "outputLimitReached": false,
          "eosReached": true
        },
        {
          "page": 2,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 95,
          "inferenceMs": 4866.91320799946,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 253,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "application/pdf",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 11417.920667001454,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 0,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 3134.4827079992683,
      "inferenceMs": 2996.0566250010743,
      "pagesPerSecond": 0.3337720628025985,
      "tokensPerSecond": 1.335088251210394,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 4,
          "inferenceMs": 2996.0566250010743,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 4,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3134.4827079992683,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6000.5071249979665,
      "inferenceMs": 5858.795374999318,
      "pagesPerSecond": 0.17068355113872133,
      "tokensPerSecond": 26.96800107991797,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 5858.795374999318,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6000.5071249979665,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6185.75141600013,
      "inferenceMs": 6042.393957999593,
      "pagesPerSecond": 0.16549731893533504,
      "tokensPerSecond": 26.148576391782935,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6042.393957999593,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6185.75141600013,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6095.4652079999505,
      "inferenceMs": 5951.139791999594,
      "pagesPerSecond": 0.1680350378165118,
      "tokensPerSecond": 26.549535975008865,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 5951.139791999594,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6095.4652079999505,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 234.13920899838558,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6593.717832998664,
      "inferenceMs": 6079.362917000253,
      "pagesPerSecond": 0.16449092012645153,
      "tokensPerSecond": 25.98956537997934,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6079.362917000253,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6827.857041997049,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6196.763749998354,
      "inferenceMs": 6047.7085419988725,
      "pagesPerSecond": 0.1653518837846446,
      "tokensPerSecond": 26.125597637973847,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6047.7085419988725,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6196.763749998354,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6067.744291001873,
      "inferenceMs": 5925.4142090030655,
      "pagesPerSecond": 0.16876457319736424,
      "tokensPerSecond": 26.66480256518355,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 5925.4142090030655,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6067.744291001873,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 231.2269579997519,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6541.495584002405,
      "inferenceMs": 6049.515958999109,
      "pagesPerSecond": 0.1653024815171245,
      "tokensPerSecond": 26.117792079705673,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 6049.515958999109,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6772.722542002157,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6077.96750000125,
      "inferenceMs": 5928.538042000582,
      "pagesPerSecond": 0.16867564868699914,
      "tokensPerSecond": 26.650752492545866,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 5928.538042000582,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6077.96750000125,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 6130.192207998334,
      "inferenceMs": 5990.318083000602,
      "pagesPerSecond": 0.1669360434862069,
      "tokensPerSecond": 26.375894870820687,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 5990.318083000602,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6130.192207998334,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 399.5997919992078,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 3308.3390409992717,
      "inferenceMs": 2836.1237499993877,
      "pagesPerSecond": 0.35259392330825334,
      "tokensPerSecond": 55.70983988270403,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2836.1237499993877,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3707.9388329984795,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 1,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2410.981000000902,
      "inferenceMs": 2261.7572499984817,
      "pagesPerSecond": 0.442134097282399,
      "tokensPerSecond": 51.287555284758284,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 116,
          "inferenceMs": 2261.7572499984817,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 116,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2410.981000000902,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 4961.91925000312,
      "inferenceMs": 4646.09120899695,
      "pagesPerSecond": 0.430469379535014,
      "tokensPerSecond": 54.45437651117927,
      "pageCount": 2,
      "pages": [
        {
          "page": 1,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2686.3760839987663,
          "outputLimitReached": false,
          "eosReached": true
        },
        {
          "page": 2,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 95,
          "inferenceMs": 1959.7151249981835,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 253,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "application/pdf",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4961.91925000312,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 0,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 1264.5018750008603,
      "inferenceMs": 1120.4285829990113,
      "pagesPerSecond": 0.8925156098065042,
      "tokensPerSecond": 3.570062439226017,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 4,
          "inferenceMs": 1120.4285829990113,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 4,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 1264.5018750008603,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2760.5528339990997,
      "inferenceMs": 2616.954083001474,
      "pagesPerSecond": 0.38212363239215336,
      "tokensPerSecond": 60.37553391796023,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2616.954083001474,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2760.5528339990997,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2906.097792001674,
      "inferenceMs": 2757.1589590006624,
      "pagesPerSecond": 0.36269218237691014,
      "tokensPerSecond": 57.305364815551805,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2757.1589590006624,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2906.097792001674,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2811.8897909989755,
      "inferenceMs": 2658.6438750018715,
      "pagesPerSecond": 0.3761316095783216,
      "tokensPerSecond": 59.428794313374816,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2658.6438750018715,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2811.8897909989755,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 463.0552919988986,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 3353.3787920023315,
      "inferenceMs": 2873.0370410012256,
      "pagesPerSecond": 0.348063733856877,
      "tokensPerSecond": 54.994069949386564,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2873.0370410012256,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3816.43408400123,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2829.662083000585,
      "inferenceMs": 2679.934499999945,
      "pagesPerSecond": 0.3731434480954742,
      "tokensPerSecond": 58.956664799084926,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2679.934499999945,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2829.662083000585,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2896.321749998606,
      "inferenceMs": 2744.022708000557,
      "pagesPerSecond": 0.36442847104886167,
      "tokensPerSecond": 57.579698425720146,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2744.022708000557,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2896.321749998606,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 452.1019579988206,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 3400.96066699698,
      "inferenceMs": 2918.955207998806,
      "pagesPerSecond": 0.34258833340768724,
      "tokensPerSecond": 54.12895667841458,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2918.955207998806,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3853.0626249958004,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2796.6863330002525,
      "inferenceMs": 2639.353167000081,
      "pagesPerSecond": 0.37888070929765394,
      "tokensPerSecond": 59.86315206902932,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2639.353167000081,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2796.6863330002525,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2915.7769170014944,
      "inferenceMs": 2767.9037079979025,
      "pagesPerSecond": 0.36128424450261176,
      "tokensPerSecond": 57.08291063141266,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 2767.9037079979025,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 15888875520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2915.7769170014944,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "phraseRecall": 0,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": false,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 398.6139170010574,
      "placement": {
        "mps:0/torch.float16": 470
      },
      "parameterDtype": "torch.float16",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 14322.553667003376,
      "inferenceMs": 13650.904040998284,
      "pagesPerSecond": 0.0732552215587086,
      "tokensPerSecond": 75.01334687611761,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 1024,
          "inferenceMs": 13650.904040998284,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 1024,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 44600983552,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp16",
      "cpuThreads": 4,
      "totalMs": 14721.167584004434,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "phraseRecall": 0,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": false,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": false,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float16": 470
      },
      "parameterDtype": "torch.float16",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 13801.189333997172,
      "inferenceMs": 13649.457999999868,
      "pagesPerSecond": 0.07326298231036057,
      "tokensPerSecond": 75.02129388580923,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 1024,
          "inferenceMs": 13649.457999999868,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 1024,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 44600983552,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp16",
      "cpuThreads": 4,
      "totalMs": 13801.189333997172,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "phraseRecall": 0,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": false,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float16": 470
      },
      "parameterDtype": "torch.float16",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 30508.169500000804,
      "inferenceMs": 30191.073416997824,
      "pagesPerSecond": 0.06624474633200632,
      "tokensPerSecond": 67.83462024397447,
      "pageCount": 2,
      "pages": [
        {
          "page": 1,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 1024,
          "inferenceMs": 14096.000292000099,
          "outputLimitReached": true,
          "eosReached": false
        },
        {
          "page": 2,
          "width": 864,
          "height": 1013,
          "inputTokens": 1142,
          "outputTokens": 1024,
          "inferenceMs": 16095.073124997725,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 2048,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "application/pdf",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 44600983552,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp16",
      "cpuThreads": 4,
      "totalMs": 30508.169500000804,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "phraseRecall": 1,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 1,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": true,
      "tableCount": 0,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 641.6284580009233,
      "placement": {
        "mps:0/torch.bfloat16": 470
      },
      "parameterDtype": "torch.bfloat16",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 5354.282874999626,
      "inferenceMs": 4728.248666997388,
      "pagesPerSecond": 0.2114947986936222,
      "tokensPerSecond": 33.41617819359231,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 158,
          "inferenceMs": 4728.248666997388,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 158,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 9706471424,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 5995.911333000549,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "phraseRecall": 0.3333333333333333,
      "minimumPhraseRecall": 0.8,
      "textTokenRecall": 0.5,
      "minimumTextTokenRecall": 0.9,
      "tableStructureAndCells": false,
      "tableCount": 1,
      "blankNoHallucination": true,
      "pageCoverage": true,
      "completeGeneration": true,
      "repetitive": false,
      "scope": "Independent authored document content and table annotations; full EOS-completed pages required"
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.bfloat16": 470
      },
      "parameterDtype": "torch.bfloat16",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 3572.004499998002,
      "inferenceMs": 3413.6911669993424,
      "pagesPerSecond": 0.2929380401095295,
      "tokensPerSecond": 28.70792793073389,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 98,
          "inferenceMs": 3413.6911669993424,
          "outputLimitReached": false,
          "eosReached": true
        }
      ],
      "outputTokens": 98,
      "maxNewTokens": 1024,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": false,
      "allPagesReachedEos": true,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 9706471424,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 3572.004499998002,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "fail",
    "quality": {
      "passed": false,
      "scope": "Original receipt content with splitting disabled",
      "expectedPartial": false,
      "validDocumentShape": true,
      "fullContentMatched": false
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 229.39883299841313,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 18381.32745799885,
      "inferenceMs": 17932.495584002027,
      "pagesPerSecond": 0.05576468681203089,
      "tokensPerSecond": 57.10303929551963,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 82,
          "outputTokens": 1024,
          "inferenceMs": 17932.495584002027,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 1024,
      "maxNewTokens": 1024,
      "doImageSplitting": false,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 18610.726290997263,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "scope": "Explicit partial-output behavior, not full-page extraction quality",
      "expectedPartial": true,
      "validDocumentShape": true,
      "fullContentMatched": false
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "cpu/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 4452.096750002966,
      "inferenceMs": 4296.683958000358,
      "pagesPerSecond": 0.23273762040096427,
      "tokensPerSecond": 3.7238019264154283,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 16,
          "inferenceMs": 4296.683958000358,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 16,
      "maxNewTokens": 16,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "cpu"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4452.096750002966,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "scope": "Original receipt content with splitting disabled",
      "expectedPartial": false,
      "validDocumentShape": true,
      "fullContentMatched": false
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 402.03674999793293,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 11528.050332999555,
      "inferenceMs": 11111.59483299707,
      "pagesPerSecond": 0.08999608202329272,
      "tokensPerSecond": 92.15598799185175,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 82,
          "outputTokens": 1024,
          "inferenceMs": 11111.59483299707,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 1024,
      "maxNewTokens": 1024,
      "doImageSplitting": false,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 1884651520,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 11930.087082997488,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scope": "Explicit partial-output behavior, not full-page extraction quality",
      "expectedPartial": true,
      "validDocumentShape": true,
      "fullContentMatched": false
    },
    "runtime": {
      "engine": "transformers",
      "version": "5.22.1",
      "torchVersion": "2.10.0",
      "doclingCoreVersion": "2.105.0",
      "pdfiumVersion": "5.14.0",
      "loadMs": 0,
      "placement": {
        "mps:0/torch.float32": 470
      },
      "parameterDtype": "torch.float32",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "attentionImplementation": "sdpa",
      "accelerationDetails": "Vision/language model parameters and input tensors on selected device; image/PDF rendering, tokenization, decoding and Docling Core conversion on CPU",
      "durationMs": 2094.0821250005683,
      "inferenceMs": 1938.2313329988392,
      "pagesPerSecond": 0.5159342865708377,
      "tokensPerSecond": 8.254948585133404,
      "pageCount": 1,
      "pages": [
        {
          "page": 1,
          "width": 768,
          "height": 900,
          "inputTokens": 1142,
          "outputTokens": 16,
          "inferenceMs": 1938.2313329988392,
          "outputLimitReached": true,
          "eosReached": false
        }
      ],
      "outputTokens": 16,
      "maxNewTokens": 16,
      "doImageSplitting": true,
      "pdfScale": 1.5,
      "inputKind": "image",
      "activationPlacement": {
        "visionOutput": "mps:0"
      },
      "decoding": "greedy",
      "outputLimitReached": true,
      "allPagesReachedEos": false,
      "structuredImagesIncluded": false,
      "driverAllocatedMemoryBytes": 6772047872,
      "memoryMeasurementScope": "Post-inference Metal driver allocation, not peak",
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 2094.0821250005683,
      "artifactRevision": "982fe3b40f2fa73c365bdb1bcacf6c81b7184bfe",
      "fallback": false,
      "grade": "fast"
    }
  }
]

Something not quite right?

Tell us what happened with Granite Docling Desktop. 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.