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.

LightOnOCR-2 1B Base

Read one still document-page image using the exact 1B Base checkpoint converted for browser OCR.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Read one still document-page image using the exact 1B Base checkpoint converted for browser OCR.

Created by
LightOn AI
License
Apache-2.0
Model released
Jan 19, 2026
Source repository created
Jan 16, 2026
Model weights · Q4
1.3 GB

Best for

  • Local document OCR with manual verification
  • Reading bounded scientific excerpts and simple scanned pages
  • Reproducible exact-checkpoint OCR experiments

Keep in mind

  • OCR can omit, repeat, misread or hallucinate text and coordinates. Verify extracted names, numbers, formulas and layout; unsuitable as legal, medical, financial, identity or redaction ground truth.
  • Quality evidence is limited to two real English scientific crops and one authored receipt; no broad multilingual, handwriting, table, low-quality scan or full-page accuracy claim.
  • Accepts one still PNG/JPEG URL, at most32MiB and16million decoded pixels. No PDF, audio, video, text chat or sampled-frame input; callers must rasterize PDF pages separately.
  • The oriented image is composited onto white and resized aspect-preservingly to1120pixels on its longest edge, then the pinned original processor normalization is applied. Coordinates use the original oriented dimensions.
  • Output is bounded to1–2048tokens and16384combined input/output tokens. Inspect eosReached/outputLimitReached. Uniform-color pages return explicit empty preprocessing-only results with generationSkipped.
  • Q4 denotes4-bit learned MatMul weights; embeddings, convolutions, norms, KV caches and graph I/O remainFP32. Browser memory is substantially greater than file size. WebGPU neural execution includes CPU preprocessing and may use WASM for unsupported shape operations.
  • Embedded-image localization is unavailable in this browser adapter; the exact checkpoint may contain localization training, but it has not passed the independent localization gate.
  • Qualified scope: Chrome157.0.8103.39 on the measured Apple Silicon Mac only. WASM is unavailable: larger-page/reused-session inference hit a runtime integer-overflow error. No CPU fallback or other browser/hardware support is advertised.
  • Benchmark complete; public artifact publication, complete dependency backup and forced original-model/runtime source recovery remain release gates. A queued upload is not a verified backup.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
https://huggingface.co/lightonai/LightOnOCR-2-1B-base
Release verification
Verified
Release evidence
Official provider announcement published2026-01-19 explicitly lists and announces this exact checkpoint among six released models. Separate from repository creation, Jan26 paper update and later repository changes.
Source updated
Jun 29, 2026
Browser model files
localai-tools/lightonocr-2-1b-base-onnx
Files updated
Nov 8, 2026
Artifact revision
d722adb98a1a6b7cec6774ac3225cb93120e0936

Input

IMAGE

Sample result

LightOnOCR-2 1B Base

Settings

Limits generated tokens for one page; combined image/prompt and output must fit16384tokens. Lower values reduce runtime but can truncate extraction; higher values allow longer text at greater time and memory cost.

Output

Your result will appear here

Add input or try the sample, then run the model.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

LightOnOCR-2 1B Base

9 configurations tested
Passed7
Failed2
TestedNov 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4CPU · WASMPassedVery slow
Q4CPU · WASMFailed—
Q4CPU · WASMFailed—
Test setup & sample details
Quality gate
Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.
Tested
2026-11-08T15:05:30.974Z
Environment
Isolated persistent Chrome157.0.8103.39 module worker per precision/device configuration on macOS Apple Silicon M3 Max96GiB. 1120px longest edge; original LightOn flat image-token expansion. Actual GPU dispatch tracing; CPU preprocessing/possible WASM shape fallback. Concurrent tracked mirror uploads recorded.
Runtime
Transformers.js 4.2.0
Artifact revision
d722adb98a1a6b7cec6774ac3225cb93120e0936
Fixture
{
  "file": "model-samples/desktop/lightonocr2/fixtures/scientific-paragraph.png",
  "url": "https://assets.localai.tools/samples/v1/files/de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb/scientific-paragraph.png",
  "size": 26674,
  "sha1": "b4e9dd3ea6464bc8e1ff1f95134eccfc25622d52",
  "sha256": "de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb"
}
Rerun when
checkpoint/artifact revision; runtime version; adapter/preprocessing/quantization; fixture checksum or quality gates; browser/backend behavior
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 5376.519999999553,
      "totalMs": 12247.519999999553,
      "observedBrowserRssBytes": 4834541568,
      "pagesPerSecond": 0.18599391427914025,
      "tokensPerSecond": 14.50752531377294,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.18599391427914025
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "phraseRecall": 0.8333333333333334,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 24769.210000000894,
      "totalMs": 31640.210000000894,
      "observedBrowserRssBytes": 4879319040,
      "pagesPerSecond": 0.04037270466034096,
      "tokensPerSecond": 12.959638195969449,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.04037270466034096
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "phraseRecall": 1,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 29176.455000001937,
      "totalMs": 36047.45500000194,
      "observedBrowserRssBytes": 4899995648,
      "pagesPerSecond": 0.03427421186021172,
      "tokensPerSecond": 1.713710593010586,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.03427421186021172
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 5291.744999997318,
      "totalMs": 12162.744999997318,
      "observedBrowserRssBytes": 4924637184,
      "pagesPerSecond": 0.1889735805486672,
      "tokensPerSecond": 14.73993928279604,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1889735805486672
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 5245.5,
      "totalMs": 12116.5,
      "observedBrowserRssBytes": 4963876864,
      "pagesPerSecond": 0.19063959584405682,
      "tokensPerSecond": 14.869888475836431,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.19063959584405682
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6871,
      "inferenceMs": 5270.469999998808,
      "totalMs": 12141.469999998808,
      "observedBrowserRssBytes": 4977049600,
      "pagesPerSecond": 0.18973639922060578,
      "tokensPerSecond": 14.799439139207252,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.18973639922060578
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 6495,
      "inferenceMs": 31809.980000000447,
      "totalMs": 38304.98000000045,
      "observedBrowserRssBytes": 5826576384,
      "pagesPerSecond": 0.031436674905170825,
      "tokensPerSecond": 2.4520606426033247,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.031436674905170825
      },
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 6495,
      "inferenceMs": 265,
      "observedBrowserRssBytes": 5182373888,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 6495,
      "inferenceMs": 66,
      "observedBrowserRssBytes": 5262950400,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  }
]

LightOnOCR-2 1B BBox Base

9 configurations tested
Passed7
Failed2
TestedNov 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4CPU · WASMPassedVery slow
Q4CPU · WASMFailed—
Q4CPU · WASMFailed—
Test setup & sample details
Quality gate
Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.
Tested
2026-11-08T15:17:25.578Z
Environment
Isolated persistent Chrome157.0.8103.39 module worker per precision/device configuration on macOS Apple Silicon M3 Max96GiB. 1344px longest edge; original LightOn flat image-token expansion. Actual GPU dispatch tracing; CPU preprocessing/possible WASM shape fallback. Concurrent tracked mirror uploads recorded.
Runtime
Transformers.js 4.2.0
Artifact revision
7827bf8c966e44575ca7155d7f82227db8c961e8
Fixture
{
  "file": "model-samples/desktop/lightonocr2/fixtures/scientific-paragraph.png",
  "url": "https://assets.localai.tools/samples/v1/files/de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb/scientific-paragraph.png",
  "size": 26674,
  "sha1": "b4e9dd3ea6464bc8e1ff1f95134eccfc25622d52",
  "sha256": "de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb"
}
Rerun when
checkpoint/artifact revision; runtime version; adapter/preprocessing/quantization; fixture checksum or quality gates; browser/backend behavior
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 8884.830000001937,
      "totalMs": 14349.830000001937,
      "observedBrowserRssBytes": 4498604032,
      "pagesPerSecond": 0.11255139378016034,
      "tokensPerSecond": 8.779008714852507,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.11255139378016034
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 0.8333333333333334,
      "textQuality": true,
      "imageBoxIoU": 0.7465105085832833,
      "bboxQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 45674.04999999702,
      "totalMs": 51139.04999999702,
      "observedBrowserRssBytes": 4553621504,
      "pagesPerSecond": 0.02189427037891462,
      "tokensPerSecond": 7.268897765799654,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.02189427037891462
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 1,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 54873.62000000477,
      "totalMs": 60338.62000000477,
      "observedBrowserRssBytes": 4564074496,
      "pagesPerSecond": 0.018223692914735953,
      "tokensPerSecond": 0.9111846457367977,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.018223692914735953
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 8622.370000004768,
      "totalMs": 14087.370000004768,
      "observedBrowserRssBytes": 4581244928,
      "pagesPerSecond": 0.11597739368635851,
      "tokensPerSecond": 9.046236707535963,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.11597739368635851
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 8661.234999999404,
      "totalMs": 14126.234999999404,
      "observedBrowserRssBytes": 4587094016,
      "pagesPerSecond": 0.1154569758238945,
      "tokensPerSecond": 9.00564411426377,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1154569758238945
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5465,
      "inferenceMs": 8687.57499999553,
      "totalMs": 14152.57499999553,
      "observedBrowserRssBytes": 4667719680,
      "pagesPerSecond": 0.11510691994031874,
      "tokensPerSecond": 8.978339755344862,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.11510691994031874
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5431,
      "inferenceMs": 39448.59499999881,
      "totalMs": 44879.59499999881,
      "observedBrowserRssBytes": 6042025984,
      "pagesPerSecond": 0.02534944527175252,
      "tokensPerSecond": 1.9772567311966969,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.02534944527175252
      },
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5431,
      "inferenceMs": 366,
      "observedBrowserRssBytes": 6141378560,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5431,
      "inferenceMs": 80,
      "observedBrowserRssBytes": 5266178048,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  }
]

LightOnOCR-2 1B BBox Soup

9 configurations tested
Passed7
Failed2
TestedNov 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4CPU · WASMPassedVery slow
Q4CPU · WASMFailed—
Q4CPU · WASMFailed—
Test setup & sample details
Quality gate
Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.
Tested
2026-11-08T15:09:56.939Z
Environment
Isolated persistent Chrome157.0.8103.39 module worker per precision/device configuration on macOS Apple Silicon M3 Max96GiB. 1120px longest edge; original LightOn flat image-token expansion. Actual GPU dispatch tracing; CPU preprocessing/possible WASM shape fallback. Concurrent tracked mirror uploads recorded.
Runtime
Transformers.js 4.2.0
Artifact revision
13b93c6793da310fd6057225f06453b68bae0ff6
Fixture
{
  "file": "model-samples/desktop/lightonocr2/fixtures/scientific-paragraph.png",
  "url": "https://assets.localai.tools/samples/v1/files/de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb/scientific-paragraph.png",
  "size": 26674,
  "sha1": "b4e9dd3ea6464bc8e1ff1f95134eccfc25622d52",
  "sha256": "de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb"
}
Rerun when
checkpoint/artifact revision; runtime version; adapter/preprocessing/quantization; fixture checksum or quality gates; browser/backend behavior
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 6175.335000000894,
      "totalMs": 11600.335000000894,
      "observedBrowserRssBytes": 4379918336,
      "pagesPerSecond": 0.16193453472562302,
      "tokensPerSecond": 10.849613826616743,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.16193453472562302
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 0.8333333333333334,
      "textQuality": true,
      "imageBoxIoU": 0.6699083507138851,
      "bboxQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 28914.734999999404,
      "totalMs": 34339.734999999404,
      "observedBrowserRssBytes": 4434345984,
      "pagesPerSecond": 0.03458444284549108,
      "tokensPerSecond": 11.51661946754853,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.03458444284549108
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 1,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 33842.224999997765,
      "totalMs": 39267.224999997765,
      "observedBrowserRssBytes": 4454711296,
      "pagesPerSecond": 0.02954888456654567,
      "tokensPerSecond": 1.4774442283272835,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.02954888456654567
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 5186.625,
      "totalMs": 10611.625,
      "observedBrowserRssBytes": 4500668416,
      "pagesPerSecond": 0.1928036054274215,
      "tokensPerSecond": 12.91784156363724,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1928036054274215
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 5075.689999997616,
      "totalMs": 10500.689999997616,
      "observedBrowserRssBytes": 4529668096,
      "pagesPerSecond": 0.19701754835312435,
      "tokensPerSecond": 13.20017573965933,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.19701754835312435
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5425,
      "inferenceMs": 5214.989999998361,
      "totalMs": 10639.98999999836,
      "observedBrowserRssBytes": 4555227136,
      "pagesPerSecond": 0.19175492186951734,
      "tokensPerSecond": 12.847579765257663,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.19175492186951734
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5411,
      "inferenceMs": 29413.75,
      "totalMs": 34824.75,
      "observedBrowserRssBytes": 5832343552,
      "pagesPerSecond": 0.033997705154902046,
      "tokensPerSecond": 2.277846245378437,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.033997705154902046
      },
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5411,
      "inferenceMs": 291,
      "observedBrowserRssBytes": 5201559552,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5411,
      "inferenceMs": 69,
      "observedBrowserRssBytes": 5282070528,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  }
]

LightOnOCR-2 1B BBox

9 configurations tested
Passed7
Failed2
TestedNov 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4CPU · WASMPassedVery slow
Q4CPU · WASMFailed—
Q4CPU · WASMFailed—
Test setup & sample details
Quality gate
Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.
Tested
2026-11-08T15:15:02.005Z
Environment
Isolated persistent Chrome157.0.8103.39 module worker per precision/device configuration on macOS Apple Silicon M3 Max96GiB. 1120px longest edge; original LightOn flat image-token expansion. Actual GPU dispatch tracing; CPU preprocessing/possible WASM shape fallback. Concurrent tracked mirror uploads recorded.
Runtime
Transformers.js 4.2.0
Artifact revision
e01aa6cd4041bdbbc9c2c32cd781e3200663f5a4
Fixture
{
  "file": "model-samples/desktop/lightonocr2/fixtures/scientific-paragraph.png",
  "url": "https://assets.localai.tools/samples/v1/files/de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb/scientific-paragraph.png",
  "size": 26674,
  "sha1": "b4e9dd3ea6464bc8e1ff1f95134eccfc25622d52",
  "sha256": "de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb"
}
Rerun when
checkpoint/artifact revision; runtime version; adapter/preprocessing/quantization; fixture checksum or quality gates; browser/backend behavior
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 5234.569999996573,
      "totalMs": 11010.569999996573,
      "observedBrowserRssBytes": 4582703104,
      "pagesPerSecond": 0.19103765925389377,
      "tokensPerSecond": 12.799523170010882,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.19103765925389377
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 0.8333333333333334,
      "textQuality": true,
      "imageBoxIoU": 0.6629013316219556,
      "bboxQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 25156.08999999985,
      "totalMs": 30932.08999999985,
      "observedBrowserRssBytes": 4634804224,
      "pagesPerSecond": 0.0397518056263913,
      "tokensPerSecond": 13.197599467961911,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.0397518056263913
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "phraseRecall": 1,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 29315.675000000745,
      "totalMs": 35091.675000000745,
      "observedBrowserRssBytes": 4657184768,
      "pagesPerSecond": 0.03411144379244123,
      "tokensPerSecond": 1.7055721896220615,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.03411144379244123
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 5173.080000001937,
      "totalMs": 10949.080000001937,
      "observedBrowserRssBytes": 4700684288,
      "pagesPerSecond": 0.1933084352068063,
      "tokensPerSecond": 12.951665158856022,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1933084352068063
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 5131.760000001639,
      "totalMs": 10907.76000000164,
      "observedBrowserRssBytes": 4709826560,
      "pagesPerSecond": 0.1948649196376449,
      "tokensPerSecond": 13.055949615722207,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1948649196376449
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5776,
      "inferenceMs": 5223.995000001043,
      "totalMs": 10999.995000001043,
      "observedBrowserRssBytes": 4722556928,
      "pagesPerSecond": 0.19142437923462796,
      "tokensPerSecond": 12.825433408720073,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.19142437923462796
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5659,
      "inferenceMs": 29402.81500000134,
      "totalMs": 35061.81500000134,
      "observedBrowserRssBytes": 5710856192,
      "pagesPerSecond": 0.034010349009098426,
      "tokensPerSecond": 2.278693383609595,
      "grade": "B",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.034010349009098426
      },
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5659,
      "inferenceMs": 262,
      "observedBrowserRssBytes": 5805981696,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5659,
      "inferenceMs": 60,
      "observedBrowserRssBytes": 5877940224,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  }
]

LightOnOCR-2 1B OCR Soup

9 configurations tested
Passed7
Failed2
TestedNov 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4GPU · WebGPUAcceleratedPassedVery slow
Q4CPU · WASMPassedVery slow
Q4CPU · WASMFailed—
Q4CPU · WASMFailed—
Test setup & sample details
Quality gate
Frozen browser-gates.json: paragraph token recall>=0.9/CER<=0.1, phrase recall>=0.8, required image-boxIoU>=0.6, EOS/non-repetition, per-page WebGPU<=60s/WASM<=120s and observed browser RSS<=16GiB. All three distinct image fixtures must pass; no sibling/device evidence transfer.
Tested
2026-11-08T15:08:24.312Z
Environment
Isolated persistent Chrome157.0.8103.39 module worker per precision/device configuration on macOS Apple Silicon M3 Max96GiB. 1120px longest edge; original LightOn flat image-token expansion. Actual GPU dispatch tracing; CPU preprocessing/possible WASM shape fallback. Concurrent tracked mirror uploads recorded.
Runtime
Transformers.js 4.2.0
Artifact revision
57782483f7e9318e0f38d69aa9e0728f7a9ba71a
Fixture
{
  "file": "model-samples/desktop/lightonocr2/fixtures/scientific-paragraph.png",
  "url": "https://assets.localai.tools/samples/v1/files/de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb/scientific-paragraph.png",
  "size": 26674,
  "sha1": "b4e9dd3ea6464bc8e1ff1f95134eccfc25622d52",
  "sha256": "de0959e7c3fe59f4d706bb84d993cd1aae000a6b25bcef96f472d3a675871dfb"
}
Rerun when
checkpoint/artifact revision; runtime version; adapter/preprocessing/quantization; fixture checksum or quality gates; browser/backend behavior
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 6642.544999998063,
      "totalMs": 12450.544999998063,
      "observedBrowserRssBytes": 4421861376,
      "pagesPerSecond": 0.15054470839118014,
      "tokensPerSecond": 11.742487254512051,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.15054470839118014
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "phraseRecall": 0.8333333333333334,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 32450.820000000298,
      "totalMs": 38258.8200000003,
      "observedBrowserRssBytes": 4461379584,
      "pagesPerSecond": 0.030815862280213283,
      "tokensPerSecond": 9.891891791948463,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.030815862280213283
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "phraseRecall": 1,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 38387.384999997914,
      "totalMs": 44195.384999997914,
      "observedBrowserRssBytes": 4481843200,
      "pagesPerSecond": 0.026050224572474898,
      "tokensPerSecond": 1.302511228623745,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.026050224572474898
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 5886.894999999553,
      "totalMs": 11694.894999999553,
      "observedBrowserRssBytes": 4493672448,
      "pagesPerSecond": 0.16986883577846656,
      "tokensPerSecond": 13.249769190720393,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.16986883577846656
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 5903,
      "totalMs": 11711,
      "observedBrowserRssBytes": 4546871296,
      "pagesPerSecond": 0.1694053870913095,
      "tokensPerSecond": 13.213620193122141,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.1694053870913095
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5808,
      "inferenceMs": 6436.094999998808,
      "totalMs": 12244.094999998808,
      "observedBrowserRssBytes": 4546641920,
      "pagesPerSecond": 0.15537371651602178,
      "tokensPerSecond": 12.1191498882497,
      "grade": "A",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.15537371651602178
      },
      "device": "webgpu",
      "precision": "q4",
      "provider": "WebGPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "complete": true,
      "nonRepetitive": true,
      "withinSpeedBudget": true,
      "withinMemoryBudget": true,
      "actualNeuralPlacement": true,
      "tokenRecall": 0.95,
      "characterErrorRate": 0.006968641114982578,
      "textQuality": true,
      "passed": true
    },
    "runtime": {
      "loadMs": 5975,
      "inferenceMs": 32193.38500000164,
      "totalMs": 38168.38500000164,
      "observedBrowserRssBytes": 5496864768,
      "pagesPerSecond": 0.031062281894244707,
      "tokensPerSecond": 2.422857987751087,
      "grade": "C",
      "performanceMetric": {
        "name": "pagesPerSecond",
        "value": 0.031062281894244707
      },
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5975,
      "inferenceMs": 266,
      "observedBrowserRssBytes": 5563006976,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "error": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
    },
    "runtime": {
      "loadMs": 5975,
      "inferenceMs": 59,
      "observedBrowserRssBytes": 5663424512,
      "grade": "F",
      "device": "wasm",
      "precision": "q4",
      "provider": "WASM CPU"
    },
    "failure": "Error: page.evaluate: Error: failed to call OrtRun(). ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/common/safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow\n\n    at worker.onmessage (eval at evaluate (:291:30), <anonymous>:1:698)"
  }
]

Something not quite right?

Tell us what happened with LightOnOCR 2. 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.