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.

LFM2.5 350M — Edge Chat

Ask questions, draft short text, and experiment with multilingual conversations using a compact chat model. Returns an assistant reply; complex reasoning and factual answers need review.

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

Ask questions, draft short text, and experiment with multilingual conversations using a compact chat model. Returns an assistant reply; complex reasoning and factual answers need review.

Created by
Liquid AI
Model released
Unknown
Source repository created
Mar 31, 2026
Model weights · Q4F16
255.1 MB

Best for

  • Local multilingual chat, drafting, rewriting, and summarization.
  • Benchmarking browser LLM inference on resource-constrained devices.

Keep in mind

  • Generated text may be incorrect, unsafe, biased, or fabricated and requires review.
  • The 350M parameter size limits reasoning depth, factual recall, and instruction following.
  • Accuracy varies by prompt language and input quality; specialized domains may perform poorly.
  • Use is governed by the custom LFM 1.0 license, and quantized artifacts may reduce output quality.

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
LiquidAI/LFM2.5-350M
Release verification
Unknown
Release evidence
Exact original checkpoint public release date has not been verified. Repository creation, training, paper, artifact conversion and update dates are not release evidence.
Source updated
Aug 5, 2026
Browser model files
onnx-community/LFM2.5-350M-ONNX
Files updated
Mar 31, 2026
Artifact revision
2c07371c2e84776cad597f3d813b7d306d292aea

Input

CHAT & TEXT

Sample result

LFM2.5 350M — Edge Chat

Settings

Limits the number of newly generated output tokens. Lower values finish sooner but can cut answers short. Higher values allow longer answers and can use more time and memory; they do not guarantee better answers.
Advanced settings 2
Adjusts randomness during token sampling. Lower values favor likely tokens and more repeatable wording. Higher values increase variation and can increase mistakes. It affects sampling only; some runtimes use zero for greedy decoding.
Chooses sampled rather than greedy token selection. On allows varied responses and makes sampling controls relevant. Off favors the most likely tokens for more repeatable responses; it does not guarantee correctness.
Estimated speedNot benchmarked

Output

Start a conversation

Send a message or try a sample. Replies appear here.

Benchmark evidence

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

LFM2.5 350M — Edge Chat

10 configurations tested
Passed7
Failed3
TestedSep 9, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

Exact-match accepted answer, non-empty/non-repetitive output, and streamed text equal to final text.

PrecisionDeviceResult
Q4F16GPU · WebGPUAcceleratedPassedFast
Q4GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedSlow
FP16GPU · WebGPUAcceleratedPassedSlow
FP32GPU · WebGPUAcceleratedPassedVery slow
FP32CPU · WASMPassedVery slow
FP16CPU · WASMPassedVery slow
Q8CPU · WASMFailed—
Q4CPU · WASMFailed—
Q4F16CPU · WASMFailed—
Test setup & sample details
Quality gate
Exact-match accepted answer, non-empty/non-repetitive output, and streamed text equal to final text.
Tested
2026-09-09T17:38:51.354Z
Environment
Not recorded
Runtime
Transformers.js 4.2.0
Artifact revision
2c07371c2e84776cad597f3d813b7d306d292aea
Fixture
{
  "file": "lfm2.5-350m.json",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/lfm2.5-350m.json",
  "size": 435,
  "sha1": "86006415fb2c4eb6d65678c2ce856f83268c2e98"
}
Rerun when
artifact revision change; runtime version change; fixture or quality gate change; browser/backend change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 103738.89999997616,
      "firstChunkMs": 319.5,
      "outputTokens": 1,
      "loadMs": 101712.6,
      "inferenceMs": 319.6,
      "tokensPerSecond": 3.129,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 111227,
      "firstChunkMs": 210.69999992847443,
      "outputTokens": 1,
      "loadMs": 109248.9,
      "inferenceMs": 210.9,
      "tokensPerSecond": 4.742,
      "grade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 64422.5,
      "firstChunkMs": 322.89999997615814,
      "outputTokens": 1,
      "loadMs": 62394.2,
      "inferenceMs": 323,
      "tokensPerSecond": 3.096,
      "grade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 57027.10000002384,
      "firstChunkMs": 139.80000007152557,
      "outputTokens": 1,
      "loadMs": 55272.7,
      "inferenceMs": 140,
      "tokensPerSecond": 7.143,
      "grade": "slow"
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 46038.799999952316,
      "loadMs": null,
      "inferenceMs": null,
      "grade": "failed"
    },
    "failure": "missing-GatherBlockQuantized-kernel"
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 45386.89999997616,
      "firstChunkMs": 108.5,
      "outputTokens": 1,
      "loadMs": 42833.9,
      "inferenceMs": 108.6,
      "tokensPerSecond": 9.208,
      "grade": "slow"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 33884.299999952316,
      "loadMs": null,
      "inferenceMs": null,
      "grade": "failed"
    },
    "failure": "missing-GatherBlockQuantized-kernel"
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 50822.60000002384,
      "firstChunkMs": 94.30000007152557,
      "outputTokens": 7,
      "loadMs": 48884.8,
      "inferenceMs": 130,
      "tokensPerSecond": 53.846,
      "grade": "excellent"
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 25498.899999976158,
      "loadMs": null,
      "inferenceMs": null,
      "grade": "failed"
    },
    "failure": "missing-GatherBlockQuantized-kernel"
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": "Not embedded",
    "runtime": {
      "totalMs": 35687.5,
      "firstChunkMs": 92.80000007152557,
      "outputTokens": 7,
      "loadMs": 33901.7,
      "inferenceMs": 128.8,
      "tokensPerSecond": 54.348,
      "grade": "excellent"
    }
  }
]

LFM2.5 230M — Text Chat

28 configurations tested
Passed21
Failed7
TestedNov 6, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

All three frozen short instruction-following fixtures; finite complete output, repetition rejection, >=5 output tokens/s.

PrecisionDeviceResult
Q4GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
UINT8GPU · WebGPUAcceleratedPassedVery fast
Q4F32GPU · WebGPUAcceleratedPassedVery fast
Q4F32GPU · WebGPUAcceleratedPassedVery fast
UINT8GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
FP16GPU · WebGPUAcceleratedPassedVery fast
FP16GPU · WebGPUAcceleratedPassedVery fast
UINT8GPU · WebGPUAcceleratedPassedVery fast
Q4F32GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedFast
FP32GPU · WebGPUAcceleratedPassedFast
FP16GPU · WebGPUAcceleratedPassedFast
FP32GPU · WebGPUAcceleratedPassedFast
FP32CPU · WASMPassedFast
FP16CPU · WASMPassedFast
FP32CPU · WASMPassedFast
FP16CPU · WASMPassedFast
FP32CPU · WASMPassedFast
FP16CPU · WASMPassedFast
Q4CPU · WASMFailed—
Q4F32CPU · WASMFailed—
Q4F32CPU · WASMFailed—
Q4F32CPU · WASMFailed—
UINT8CPU · WASMFailed—
UINT8CPU · WASMFailed—
UINT8CPU · WASMFailed—
Test setup & sample details
Quality gate
All three frozen short instruction-following fixtures; finite complete output, repetition rejection, >=5 output tokens/s.
Tested
2026-11-06T10:18:52.591Z
Environment
Isolated Chrome module worker; verified local artifact mirror; Apple M3 Max/macOS 27.2
Runtime
Transformers.js 4.2.0
Artifact revision
c6f46e4e3f885ebcad164d14059a49f90e27eb4d
Fixture
{
  "file": "model-samples/lfm25-extract.json",
  "url": "https://assets.localai.tools/samples/v1/files/614a2f1c76c698094d528a8f230e95abe459b001111deb590623151e5633958b/lfm25-extract.json",
  "size": 348,
  "sha1": "702b94d38d786b6847a93d562bb24e5564dc9168",
  "sha256": "614a2f1c76c698094d528a8f230e95abe459b001111deb590623151e5633958b"
}
Rerun when
artifact revision change; runtime version change; fixture or quality gate change; browser/backend change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 1725.199999988079,
      "inferenceMs": 508.4000000357628,
      "totalMs": 3098.699999988079,
      "grade": "excellent",
      "outputTokens": 27,
      "tokensPerSecond": 53.10778913867175
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 1725.199999988079,
      "inferenceMs": 535,
      "totalMs": 3098.699999988079,
      "grade": "excellent",
      "outputTokens": 39,
      "tokensPerSecond": 72.89719626168224
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1725.199999988079,
      "inferenceMs": 211.60000002384186,
      "totalMs": 3098.699999988079,
      "grade": "excellent",
      "outputTokens": 15,
      "tokensPerSecond": 70.88846880108642
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 2035.4000000357628,
      "inferenceMs": 897.7999999523163,
      "totalMs": 4611,
      "grade": "excellent",
      "outputTokens": 27,
      "tokensPerSecond": 30.073513033452905
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 2035.4000000357628,
      "inferenceMs": 904.2999999523163,
      "totalMs": 4611,
      "grade": "excellent",
      "outputTokens": 39,
      "tokensPerSecond": 43.12728077193019
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 2035.4000000357628,
      "inferenceMs": 577.1999999880791,
      "totalMs": 4611,
      "grade": "good",
      "outputTokens": 15,
      "tokensPerSecond": 25.98752598806271
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 1338.300000011921,
      "inferenceMs": 399.4000000357628,
      "totalMs": 2435.599999964237,
      "grade": "excellent",
      "outputTokens": 27,
      "tokensPerSecond": 67.60140209710161
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 1338.300000011921,
      "inferenceMs": 438.4000000357628,
      "totalMs": 2435.599999964237,
      "grade": "excellent",
      "outputTokens": 39,
      "tokensPerSecond": 88.95985400734158
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1338.300000011921,
      "inferenceMs": 162.70000004768372,
      "totalMs": 2435.599999964237,
      "grade": "excellent",
      "outputTokens": 15,
      "tokensPerSecond": 92.19422246837023
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 1639.300000011921,
      "inferenceMs": 936.8000000119209,
      "totalMs": 4253.200000047684,
      "grade": "good",
      "outputTokens": 27,
      "tokensPerSecond": 28.82152006795092
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 1639.300000011921,
      "inferenceMs": 981,
      "totalMs": 4253.200000047684,
      "grade": "excellent",
      "outputTokens": 39,
      "tokensPerSecond": 39.75535168195719
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1639.300000011921,
      "inferenceMs": 594.8999999761581,
      "totalMs": 4253.200000047684,
      "grade": "good",
      "outputTokens": 15,
      "tokensPerSecond": 25.21432173575585
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 860.5,
      "inferenceMs": 205.5,
      "totalMs": 1426.699999988079,
      "grade": "excellent",
      "outputTokens": 20,
      "tokensPerSecond": 97.32360097323601
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you add the numbers together:\n\n7 + 5 = 12\n\nSo, the answer is 12."
    },
    "runtime": {
      "loadMs": 860.5,
      "inferenceMs": 170.5999999642372,
      "totalMs": 1426.699999988079,
      "grade": "excellent",
      "outputTokens": 35,
      "tokensPerSecond": 205.15826499025223
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 860.5,
      "inferenceMs": 87.10000002384186,
      "totalMs": 1426.699999988079,
      "grade": "excellent",
      "outputTokens": 16,
      "tokensPerSecond": 183.69690006452726
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 961.5999999642372,
      "grade": "failed"
    },
    "failure": "Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Could not find an implementation for GatherBlockQuantized(1) node with name '/model/embed_tokens/GatherBlockQuantized'"
  },
  {
    "precision": "q4f32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 1351.5,
      "inferenceMs": 245.19999998807907,
      "totalMs": 2000.5,
      "grade": "excellent",
      "outputTokens": 20,
      "tokensPerSecond": 81.56606851946307
    }
  },
  {
    "precision": "q4f32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you add the numbers together:\n\n7 + 5 = 12\n\nSo, the answer is 12."
    },
    "runtime": {
      "loadMs": 1351.5,
      "inferenceMs": 206.69999998807907,
      "totalMs": 2000.5,
      "grade": "excellent",
      "outputTokens": 35,
      "tokensPerSecond": 169.32752782785943
    }
  },
  {
    "precision": "q4f32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1351.5,
      "inferenceMs": 98.10000002384186,
      "totalMs": 2000.5,
      "grade": "excellent",
      "outputTokens": 16,
      "tokensPerSecond": 163.09887865557005
    }
  },
  {
    "precision": "q4f32",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 1442.5,
      "inferenceMs": 4328.300000011921,
      "totalMs": 15888.300000011921,
      "grade": "slow",
      "outputTokens": 20,
      "tokensPerSecond": 4.620751796304535
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "q4f32",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you add the numbers together:\n\n7 + 5 = 12\n\nSo, the answer is 12."
    },
    "runtime": {
      "loadMs": 1442.5,
      "inferenceMs": 6671.5,
      "totalMs": 15888.300000011921,
      "grade": "acceptable",
      "outputTokens": 35,
      "tokensPerSecond": 5.246196507532039
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "q4f32",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1442.5,
      "inferenceMs": 3363.600000023842,
      "totalMs": 15888.300000011921,
      "grade": "slow",
      "outputTokens": 16,
      "tokensPerSecond": 4.756808181676355
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 1918.5,
      "inferenceMs": 321,
      "totalMs": 3004.800000011921,
      "grade": "excellent",
      "outputTokens": 27,
      "tokensPerSecond": 84.11214953271028
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 1918.5,
      "inferenceMs": 224,
      "totalMs": 3004.800000011921,
      "grade": "excellent",
      "outputTokens": 39,
      "tokensPerSecond": 174.10714285714286
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1918.5,
      "inferenceMs": 97.10000002384186,
      "totalMs": 3004.800000011921,
      "grade": "excellent",
      "outputTokens": 15,
      "tokensPerSecond": 154.47991757277973
    }
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "```json\n{\n  \"name\": \"Maya\",\n  \"date\": \"2026-04-17\"\n}\n```"
    },
    "runtime": {
      "loadMs": 1462.5,
      "inferenceMs": 5896.400000035763,
      "totalMs": 19024.30000001192,
      "grade": "slow",
      "outputTokens": 27,
      "tokensPerSecond": 4.5790651922929655
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "To find the sum of 7 and 5, you simply add them together:\n\n7 + 5 = 12\n\nSo, the answer is \\(\\boxed{12}\\)."
    },
    "runtime": {
      "loadMs": 1462.5,
      "inferenceMs": 8155.699999988079,
      "totalMs": 19024.30000001192,
      "grade": "slow",
      "outputTokens": 39,
      "tokensPerSecond": 4.781931655168411
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "The team meeting is on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1462.5,
      "inferenceMs": 3316.400000035763,
      "totalMs": 19024.30000001192,
      "grade": "slow",
      "outputTokens": 15,
      "tokensPerSecond": 4.522976721697698
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  }
]

LFM2.5 1.2B-Instruct — Text Chat

24 configurations tested
Passed15
Failed9
TestedNov 6, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

All three frozen short instruction-following fixtures; finite complete output, repetition rejection, >=5 output tokens/s.

PrecisionDeviceResult
Q4F16GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
Q8GPU · WebGPUAcceleratedPassedVery fast
Q4F16GPU · WebGPUAcceleratedPassedFast
UINT8GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedFast
Q4F16GPU · WebGPUAcceleratedPassedFast
Q4GPU · WebGPUAcceleratedPassedFast
UINT8GPU · WebGPUAcceleratedPassedFast
Q8GPU · WebGPUAcceleratedPassedFast
UINT8GPU · WebGPUAcceleratedPassedFast
FP16GPU · WebGPUAcceleratedPassedFast
FP16GPU · WebGPUAcceleratedPassedFast
FP16GPU · WebGPUAcceleratedPassedFast
FP32GPU · WebGPUFailed—
FP32CPU · WASMFailed—
FP16CPU · WASMFailed—
Q4CPU · WASMFailed—
Q4F16CPU · WASMFailed—
Q8CPU · WASMFailed—
UINT8CPU · WASMFailed—
UINT8CPU · WASMFailed—
UINT8CPU · WASMFailed—
Test setup & sample details
Quality gate
All three frozen short instruction-following fixtures; finite complete output, repetition rejection, >=5 output tokens/s.
Tested
2026-11-06T10:20:19.793Z
Environment
Isolated Chrome module worker; verified local artifact mirror; Apple M3 Max/macOS 27.2
Runtime
Transformers.js 4.2.0
Artifact revision
10f72e70abf67ac0fd7ebf15bc5854726891d864
Fixture
{
  "file": "model-samples/lfm25-extract.json",
  "url": "https://assets.localai.tools/samples/v1/files/614a2f1c76c698094d528a8f230e95abe459b001111deb590623151e5633958b/lfm25-extract.json",
  "size": 348,
  "sha1": "702b94d38d786b6847a93d562bb24e5564dc9168",
  "sha256": "614a2f1c76c698094d528a8f230e95abe459b001111deb590623151e5633958b"
}
Rerun when
artifact revision change; runtime version change; fixture or quality gate change; browser/backend change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": null,
      "grade": "unsupported"
    },
    "failure": "Exact selected weights exceed 4GiB documented ORT Web model-memory bound; not executed"
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": null,
      "grade": "unsupported"
    },
    "failure": "Exact selected weights exceed 4GiB documented ORT Web model-memory bound; not executed"
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 3687.399999976158,
      "inferenceMs": 705.4000000357628,
      "totalMs": 5378,
      "grade": "good",
      "outputTokens": 20,
      "tokensPerSecond": 28.35270768214634
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 3687.399999976158,
      "inferenceMs": 238.5,
      "totalMs": 5378,
      "grade": "excellent",
      "outputTokens": 9,
      "tokensPerSecond": 37.735849056603776
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 3687.399999976158,
      "inferenceMs": 405.69999998807907,
      "totalMs": 5378,
      "grade": "excellent",
      "outputTokens": 16,
      "tokensPerSecond": 39.43800838173561
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 4473.199999988079,
      "grade": "failed"
    },
    "failure": "Can't create a session. ERROR_CODE: 6, ERROR_MESSAGE: std::bad_alloc"
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 1698.0999999642372,
      "inferenceMs": 281.5,
      "totalMs": 2710.800000011921,
      "grade": "excellent",
      "outputTokens": 20,
      "tokensPerSecond": 71.04795737122558
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 1698.0999999642372,
      "inferenceMs": 95.5,
      "totalMs": 2710.800000011921,
      "grade": "excellent",
      "outputTokens": 9,
      "tokensPerSecond": 94.24083769633508
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting will take place on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1698.0999999642372,
      "inferenceMs": 168.5,
      "totalMs": 2710.800000011921,
      "grade": "excellent",
      "outputTokens": 17,
      "tokensPerSecond": 100.89020771513353
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 2275.199999988079,
      "grade": "failed"
    },
    "failure": "Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Could not find an implementation for GatherBlockQuantized(1) node with name '/model/embed_tokens/Gather_Quant'"
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "{\"name\": \"Maya\", \"date\": \"2026-04-17\"}"
    },
    "runtime": {
      "loadMs": 1802.300000011921,
      "inferenceMs": 274,
      "totalMs": 3675.899999976158,
      "grade": "excellent",
      "outputTokens": 20,
      "tokensPerSecond": 72.992700729927
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 1802.300000011921,
      "inferenceMs": 116,
      "totalMs": 3675.899999976158,
      "grade": "excellent",
      "outputTokens": 9,
      "tokensPerSecond": 77.58620689655172
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting will take place on Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 1802.300000011921,
      "inferenceMs": 161.5,
      "totalMs": 3675.899999976158,
      "grade": "excellent",
      "outputTokens": 17,
      "tokensPerSecond": 105.26315789473684
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 1977.199999988079,
      "grade": "failed"
    },
    "failure": "Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Failed to find kernel for com.microsoft.GatherBlockQuantized(1) (node:'/model/embed_tokens/Gather_Quant' ep:'CPUExecutionProvider'). Kernel not found"
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\"name\": \"Maya\", \"date\": \"2026-04-17\"}\n```"
    },
    "runtime": {
      "loadMs": 2336.199999988079,
      "inferenceMs": 433.69999998807907,
      "totalMs": 3459.199999988079,
      "grade": "excellent",
      "outputTokens": 24,
      "tokensPerSecond": 55.33779110135965
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 2336.199999988079,
      "inferenceMs": 119,
      "totalMs": 3459.199999988079,
      "grade": "excellent",
      "outputTokens": 9,
      "tokensPerSecond": 75.63025210084034
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 2336.199999988079,
      "inferenceMs": 190.10000002384186,
      "totalMs": 3459.199999988079,
      "grade": "excellent",
      "outputTokens": 16,
      "tokensPerSecond": 84.16622829033835
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 3304,
      "grade": "failed"
    },
    "failure": "Can't create a session. ERROR_CODE: 9, ERROR_MESSAGE: Could not find an implementation for GatherBlockQuantized(1) node with name '/model/embed_tokens/Gather_Quant'"
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "```json\n{\"name\": \"Maya\", \"date\": \"2026-04-17\"}\n```"
    },
    "runtime": {
      "loadMs": 4159.100000023842,
      "inferenceMs": 536.1999999880791,
      "totalMs": 5368,
      "grade": "excellent",
      "outputTokens": 24,
      "tokensPerSecond": 44.75941812855944
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 4159.100000023842,
      "inferenceMs": 131.5,
      "totalMs": 5368,
      "grade": "excellent",
      "outputTokens": 9,
      "tokensPerSecond": 68.44106463878326
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 4159.100000023842,
      "inferenceMs": 210.4000000357628,
      "totalMs": 5368,
      "grade": "excellent",
      "outputTokens": 16,
      "tokensPerSecond": 76.04562736349999
    }
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "```json\n{\"name\": \"Maya\", \"date\": \"2026-04-17\"}\n```"
    },
    "runtime": {
      "loadMs": 4390.899999976158,
      "inferenceMs": 54203.80000001192,
      "totalMs": 115692.89999997616,
      "grade": "slow",
      "outputTokens": 24,
      "tokensPerSecond": 0.44277338489173673
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "7 + 5 equals 12."
    },
    "runtime": {
      "loadMs": 4390.899999976158,
      "inferenceMs": 20254.799999952316,
      "totalMs": 115692.89999997616,
      "grade": "slow",
      "outputTokens": 9,
      "tokensPerSecond": 0.44433911961713707
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "passed": false,
      "degenerate": false,
      "answer": "The team meeting is scheduled for Friday at 10 am in room B."
    },
    "runtime": {
      "loadMs": 4390.899999976158,
      "inferenceMs": 36521.19999998808,
      "totalMs": 115692.89999997616,
      "grade": "slow",
      "outputTokens": 16,
      "tokensPerSecond": 0.43810170531103093
    },
    "failure": "Frozen quality/repetition/interactive speed gate failed"
  }
]

Something not quite right?

Tell us what happened with LFM 2.5. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

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