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 Embedding R2

Compare meaning, explore text vectors, and build semantic search on your device.

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

Represent multilingual queries, passages, documents, or code as 384-number vectors. Use these vectors for search and similarity comparisons; the model does not answer questions or execute code.

Created by
IBM Granite
License
Apache-2.0
Model released
Apr 29, 2026
Source repository created
Apr 20, 2026
Model weights · FP32
390.0 MB

Best for

  • Multilingual and cross-lingual semantic search and retrieval.
  • Code, document, passage, and conversational retrieval prototypes.
  • Clustering and similarity workflows with human-reviewed thresholds.

Keep in mind

  • Enhanced retrieval training covers 52 listed languages and code; quality varies outside them despite broader pretraining.
  • Similarity is not factual verification and can reproduce training-data bias.
  • Long inputs increase browser memory and latency and should be chunked even though the architecture supports up to 32,768 tokens.
  • Quantization may change close rankings; application-specific evaluation is required.
  • Do not use similarity alone for consequential decisions about people.

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
ibm-granite/granite-embedding-97m-multilingual-r2
Release verification
Verified
Release evidence
Exact checkpoint Release Date stated by original provider in model card.
Source updated
May 18, 2026
Browser model files
onnx-community/granite-embedding-97m-multilingual-r2-ONNX
Files updated
Jun 21, 2026
Artifact revision
536a9f241cb3f02a9c5995a1e708c784bd274859

Input

TEXT → VECTORS

Sample result

Granite Embedding 97M Multilingual R2
Advanced settings 1
L2-normalize embeddings for cosine similarity. On makes vector lengths comparable for cosine similarity. Off preserves raw magnitudes only if supported; the listed fixed value must be retained.

Output

Explore what your text means

Generate vectors from one or more texts. Add multiple texts to compare their meaning.

Benchmark evidence

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

Granite Embedding 97M Multilingual R2

16 configurations tested
Passed12
Failed4
TestedSep 15, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Three finite, distinct, normalized 384d vectors; an English query must rank its Japanese answer above an unrelated passage by at least 0.1 cosine.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedFast
BNB4GPU · WebGPUAcceleratedPassedFast
FP32GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedFast
INT8GPU · WebGPUAcceleratedPassedFast
UINT8GPU · WebGPUAcceleratedPassedFast
FP32CPU · WASMPassedVery fast
UINT8CPU · WASMPassedVery fast
Q8CPU · WASMPassedVery fast
INT8CPU · WASMPassedVery fast
Q4CPU · WASMPassedFast
BNB4CPU · WASMPassedFast
FP16CPU · WASMFailed—
FP16GPU · WebGPUFailed—
Q4F16CPU · WASMFailed—
Q4F16GPU · WebGPUFailed—
Test setup & sample details
Quality gate
Three finite, distinct, normalized 384d vectors; an English query must rank its Japanese answer above an unrelated passage by at least 0.1 cosine.
Tested
2026-09-15T16:25:44.372Z
Environment
Isolated headless Chrome on macOS; Transformers.js 4.2.0
Runtime
Transformers.js 4.2.0
Artifact revision
536a9f241cb3f02a9c5995a1e708c784bd274859
Fixture
{
  "file": "granite-embedding-97m-multilingual-r2.json",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/granite-embedding-97m-multilingual-r2.json",
  "size": 735,
  "sha1": "721d37a672010fd5ec9cdf564c697319d598e342"
}
Rerun when
artifact revision change; runtime version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9259532456887626,
      "negative": 0.6901542851041528,
      "margin": 0.2357989605846098,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 44256.80000001192,
      "inferenceMs": 88.90000000596046,
      "totalMs": 45639.09999999404,
      "grade": "excellent",
      "embeddingsPerSecond": 33.745781775015296
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9259531192134285,
      "negative": 0.6901542423153498,
      "margin": 0.23579887689807866,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 44079.79999998212,
      "inferenceMs": 122,
      "totalMs": 45423,
      "grade": "excellent",
      "embeddingsPerSecond": 24.59016393442623
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 29303.29999998212,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 27575.29999998212,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  },
  {
    "precision": "int8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8823177193527786,
      "negative": 0.5999481015019146,
      "margin": 0.282369617850864,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 18007.59999999404,
      "inferenceMs": 97.80000001192093,
      "totalMs": 19350.29999998212,
      "grade": "excellent",
      "embeddingsPerSecond": 30.674846622027886
    }
  },
  {
    "precision": "int8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8823176872969928,
      "negative": 0.5999480410167466,
      "margin": 0.2823696462802462,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 18234.59999999404,
      "inferenceMs": 184.19999998807907,
      "totalMs": 19668.100000023842,
      "grade": "good",
      "embeddingsPerSecond": 16.286644952194095
    }
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8717737596771854,
      "negative": 0.6119479442542025,
      "margin": 0.2598258154229829,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 19023.90000000596,
      "inferenceMs": 96.69999998807907,
      "totalMs": 20419.09999999404,
      "grade": "excellent",
      "embeddingsPerSecond": 31.02378490558255
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8693715912393482,
      "negative": 0.6193572420061288,
      "margin": 0.2500143492332193,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 20835.90000000596,
      "inferenceMs": 252.30000001192093,
      "totalMs": 22369,
      "grade": "good",
      "embeddingsPerSecond": 11.890606420365648
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8823177193527786,
      "negative": 0.5999481015019146,
      "margin": 0.282369617850864,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 19433,
      "inferenceMs": 97.5,
      "totalMs": 20740.79999998212,
      "grade": "excellent",
      "embeddingsPerSecond": 30.769230769230766
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8823176872969928,
      "negative": 0.5999480410167466,
      "margin": 0.2823696462802462,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 19048.29999998212,
      "inferenceMs": 183.7000000178814,
      "totalMs": 20532.20000001788,
      "grade": "good",
      "embeddingsPerSecond": 16.33097441321709
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9234198756914827,
      "negative": 0.6984499903196258,
      "margin": 0.2249698853718568,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 35297.89999997616,
      "inferenceMs": 125.40000000596046,
      "totalMs": 36728.09999999404,
      "grade": "excellent",
      "embeddingsPerSecond": 23.923444974939436
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.92342002894741,
      "negative": 0.6984500657764473,
      "margin": 0.22496996317096274,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 35796.59999999404,
      "inferenceMs": 107.40000000596046,
      "totalMs": 37176.40000000596,
      "grade": "excellent",
      "embeddingsPerSecond": 27.932960892304532
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 23773,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 24294.90000000596,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  },
  {
    "precision": "bnb4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9234198756914827,
      "negative": 0.6984499903196258,
      "margin": 0.2249698853718568,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 37582.90000000596,
      "inferenceMs": 129.59999999403954,
      "totalMs": 39018.89999997616,
      "grade": "excellent",
      "embeddingsPerSecond": 23.14814814921276
    }
  },
  {
    "precision": "bnb4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.92342002894741,
      "negative": 0.6984500657764473,
      "margin": 0.22496996317096274,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 34449.20000001788,
      "inferenceMs": 107.7999999821186,
      "totalMs": 35773.5,
      "grade": "excellent",
      "embeddingsPerSecond": 27.829313548215463
    }
  }
]

Granite Embedding Small English R2

10 configurations tested
Passed10
Failed0
TestedNov 5, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Three finite, distinct normalized 384-dimensional vectors; published English query ranks its matching answer above unrelated passage by at least 0.1 cosine. At least 3 embeddings/second.

PrecisionDeviceResult
Q4F16GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
FP16GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedModerate
FP32CPU · WASMPassedVery fast
Q8CPU · WASMPassedVery fast
Q4CPU · WASMPassedVery fast
Q4F16CPU · WASMPassedVery fast
FP16CPU · WASMPassedFast
Test setup & sample details
Quality gate
Three finite, distinct normalized 384-dimensional vectors; published English query ranks its matching answer above unrelated passage by at least 0.1 cosine. At least 3 embeddings/second.
Tested
2026-11-05T02:08:11.836Z
Environment
Isolated headless Chrome on Apple M3 Max, macOS 27.2; Transformers.js 4.2.0; remote pinned Hugging Face artifacts
Runtime
Transformers.js 4.2.0
Artifact revision
1dc7835ba0cb9c76a3618d0bf0c427c97671b3c8
Fixture
{
  "file": "model-samples/granite-embedding-small-english-r2.json",
  "url": "https://assets.localai.tools/samples/v1/files/9ca55d8731b5db28844f4ceb16b87098943a693c5ad5c6d7670d051ad949add9/granite-embedding-small-english-r2.json",
  "size": 776,
  "sha1": "2e90eabd532c67b92b105a13b780ae20b20ccec1",
  "sha256": "9ca55d8731b5db28844f4ceb16b87098943a693c5ad5c6d7670d051ad949add9"
}
Rerun when
artifact revision change; runtime version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8946836395276746,
      "negative": 0.6421105916434995,
      "margin": 0.2525730478841751,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 12426.230000019073,
      "inferenceMs": 74.80000001192093,
      "totalMs": 13729.724999964237,
      "grade": "excellent",
      "embeddingsPerSecond": 40.106951865265884
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.887667146273241,
      "negative": 0.6396001584737889,
      "margin": 0.24806698779945202,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 10960.17000001669,
      "inferenceMs": 524.289999961853,
      "totalMs": 12998.435000002384,
      "grade": "slow",
      "embeddingsPerSecond": 5.722024071064254
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.897338983034445,
      "negative": 0.6701277444583359,
      "margin": 0.22721123857610914,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 31281.375,
      "inferenceMs": 72.40499997138977,
      "totalMs": 32924.54500001669,
      "grade": "excellent",
      "embeddingsPerSecond": 41.43360266812271
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8973388828303291,
      "negative": 0.6701276171495176,
      "margin": 0.2272112656808115,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 24451.96500003338,
      "inferenceMs": 72.34999996423721,
      "totalMs": 25794.584999978542,
      "grade": "excellent",
      "embeddingsPerSecond": 41.4651002278218
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.897336201393085,
      "negative": 0.6701773948585082,
      "margin": 0.22715880653457676,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 17198.289999961853,
      "inferenceMs": 123.06000000238419,
      "totalMs": 18644.939999997616,
      "grade": "excellent",
      "embeddingsPerSecond": 24.37835202293091
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.8973075905369052,
      "negative": 0.6703583671012001,
      "margin": 0.22694922343570512,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 15884.924999952316,
      "inferenceMs": 396.8700000047684,
      "totalMs": 17807.044999957085,
      "grade": "slow",
      "embeddingsPerSecond": 7.559150351409669
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9035131341514614,
      "negative": 0.6706810051798565,
      "margin": 0.23283212897160488,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 14438.754999995232,
      "inferenceMs": 84.8299999833107,
      "totalMs": 15914.475000023842,
      "grade": "excellent",
      "embeddingsPerSecond": 35.364847348699925
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.903513017067584,
      "negative": 0.6706809752369691,
      "margin": 0.2328320418306149,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 13279.394999980927,
      "inferenceMs": 86.45999997854233,
      "totalMs": 14762.939999997616,
      "grade": "excellent",
      "embeddingsPerSecond": 34.69812630979113
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9034855975891627,
      "negative": 0.6707136448151247,
      "margin": 0.23277195277403795,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 12784.409999966621,
      "inferenceMs": 85.49500000476837,
      "totalMs": 14265.524999976158,
      "grade": "excellent",
      "embeddingsPerSecond": 35.08977132969974
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0.9033791704755302,
      "negative": 0.6701982644244359,
      "margin": 0.23318090605109432,
      "minimumMargin": 0.1,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 10973.290000021458,
      "inferenceMs": 72.12999999523163,
      "totalMs": 12542.824999988079,
      "grade": "excellent",
      "embeddingsPerSecond": 41.591570777739136
    }
  }
]

Granite Embedding English R2

4 configurations tested
Passed4
Failed0
TestedNov 5, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Finite, distinct normalized 768-dimensional vectors; every published query ranks its matching passage above other fixture passages by at least0.05 cosine. At least 3 embeddings/second.

PrecisionDeviceResult
Q4F16GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedModerate
Q8CPU · WASMPassedFast
Q4F16CPU · WASMPassedFast
Test setup & sample details
Quality gate
Finite, distinct normalized 768-dimensional vectors; every published query ranks its matching passage above other fixture passages by at least0.05 cosine. At least 3 embeddings/second.
Tested
2026-11-05T02:55:06.166Z
Environment
Isolated headless Chrome on Apple M3 Max, macOS 27.2; Transformers.js 4.2.0; remote pinned Hugging Face artifacts
Runtime
Transformers.js 4.2.0
Artifact revision
2a49b9c076aa627b14bc528b36b67462808ccc23
Fixture
{
  "file": "model-samples/granite-embedding-english-r2.json",
  "url": "https://assets.localai.tools/samples/v1/files/8781a5ac5bab021ca8a08b3c38cfdf635e7e59fa14337b0dd5b6b347bb87ba1c/granite-embedding-english-r2.json",
  "size": 734,
  "sha1": "2637dadeee3552c1d96e0ef369062a2fcac5cf2e",
  "sha256": "8781a5ac5bab021ca8a08b3c38cfdf635e7e59fa14337b0dd5b6b347bb87ba1c"
}
Rerun when
artifact revision change; runtime version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.24793940463093544,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 21239.160000026226,
      "inferenceMs": 247.30500000715256,
      "totalMs": 22851.649999976158,
      "grade": "good",
      "embeddingsPerSecond": 16.174359595981933
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.25092890480279295,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 22378.07999998331,
      "inferenceMs": 515.3249999880791,
      "totalMs": 24308.664999961853,
      "grade": "slow",
      "embeddingsPerSecond": 7.762091883942232
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.20901991339133574,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 48075.48000001907,
      "inferenceMs": 306.7249999642372,
      "totalMs": 49894.485000014305,
      "grade": "good",
      "embeddingsPerSecond": 13.040997637839702
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.20897933153705095,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 21231.865000009537,
      "inferenceMs": 164.11000001430511,
      "totalMs": 22776.549999952316,
      "grade": "excellent",
      "embeddingsPerSecond": 24.373895555732915
    }
  }
]

Granite Embedding 311M Multilingual R2

4 configurations tested
Passed2
Failed2
TestedNov 5, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Finite, distinct normalized 768-dimensional vectors; every published query ranks its matching passage above other fixture passages by at least0.05 cosine. At least 3 embeddings/second.

PrecisionDeviceResult
Q8GPU · WebGPUAcceleratedPassedModerate
Q8CPU · WASMPassedFast
Q4F16CPU · WASMFailed—
Q4F16GPU · WebGPUFailed—
Test setup & sample details
Quality gate
Finite, distinct normalized 768-dimensional vectors; every published query ranks its matching passage above other fixture passages by at least0.05 cosine. At least 3 embeddings/second.
Tested
2026-11-05T02:59:49.430Z
Environment
Isolated headless Chrome on Apple M3 Max, macOS 27.2; Transformers.js 4.2.0; remote pinned Hugging Face artifacts
Runtime
Transformers.js 4.2.0
Artifact revision
8f039f21d4181327268271bea4b11ddcc7eef88d
Fixture
{
  "file": "model-samples/granite-embedding-311m-multilingual-r2.json",
  "url": "https://assets.localai.tools/samples/v1/files/0d3ac888c92e8eb5239e89ddd8e70edf6a66f1db1dbeed9479bffeb3c2094ff3/granite-embedding-311m-multilingual-r2.json",
  "size": 1029,
  "sha1": "eba0ec260470b198e735dbdda42e0ec921e16cfd",
  "sha256": "0d3ac888c92e8eb5239e89ddd8e70edf6a66f1db1dbeed9479bffeb3c2094ff3"
}
Rerun when
artifact revision change; runtime version change; fixture or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.14364925317694932,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 52766.00499999523,
      "inferenceMs": 362.80000001192093,
      "totalMs": 54406.37999999523,
      "grade": "good",
      "embeddingsPerSecond": 16.538037485674895
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "shape": true,
      "finite": true,
      "normalized": true,
      "distinct": true,
      "positive": 0,
      "negative": 0,
      "margin": 0.13828977191483538,
      "minimumMargin": 0.05,
      "grade": "semantic-retrieval-pass"
    },
    "runtime": {
      "loadMs": 37048.47000002861,
      "inferenceMs": 685.1899999976158,
      "totalMs": 39137.905000030994,
      "grade": "slow",
      "embeddingsPerSecond": 8.756695223253226
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 126393.16000002623,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 53272.80500000715,
      "grade": "failed"
    },
    "failure": "Error: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: Type Error: Type (tensor(float16)) of output arg (/layers.0/attn/rotary_emb/Cast_1_output_0) of node (/layers.0/attn/rotary_emb/Cast_1) does not match expected type (tensor(float))."
  }
]

Something not quite right?

Tell us what happened with Granite Embedding R2. 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.