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.

LateOn Code Desktop

Encode natural-language queries and source-code snippets as 128-number vectors for each retained token.

Desktop inferencePrivate · Runs on your computer
About this model

Encode natural-language queries and source-code snippets as 128-number vectors for each retained token.

Created by
LightOn AI
License
Apache-2.0
Model released
Feb 12, 2026
Source repository created
Feb 11, 2026
Model weights · FP32
596.5 MB

Best for

  • Late-interaction text retrieval using MaxSim.
  • Token-level query and passage comparison.

Keep in mind

  • Measured on three small code retrieval fixtures. Evaluate on your own corpus before relying on rankings.
  • Queries are truncated to 256 tokens and documents to 2048 tokens, including special tokens.
  • Retrieval scores measure relevance and can reflect training-data biases. Scores do not establish factual accuracy.
  • The selected runtime returns complete token matrices after inference. Incremental streaming is unavailable.

Processed on your deviceYour input is processed locally in the desktop app. Model and runtime files download from external hosts.

Target platforms · verification required

  • macos · aarch64, x86_64
  • windows · x86_64

Recommended configuration

FP32 · cpu (verified platform only)

Sources & technical details
Original model
lightonai/LateOn-Code
Release verification
Verified
Release evidence
LightOn’s dated release announcement identifies the exact checkpoint and links the pretrained baselines.
Source updated
Aug 17, 2026
Native model files
lightonai/LateOn-Code
Files updated
Aug 17, 2026
Artifact revision
ace431824f35db231178fc602e33296784762a2e

Desktop app

Sample result

LateOn Code Base Desktop

Run LateOn Code Desktop on your computer

Download the desktop app to load this model, provide input and view results locally.

Public desktop downloads are coming soon.

Device requirements

macOS Apple SiliconCPU / FP32Tested8.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested8.0 GiBRAM estimate
macOS Apple SiliconMPS / FP16Tested8.0 GiBRAM estimate
macOS Apple SiliconMPS / BF16Tested8.0 GiBRAM estimate
If your device falls short

MemoryAdvisory only: close other apps, reduce image/tile/token/batch settings or choose a smaller verified model/precision. CPU can need more RAM than GPU.

GPUIf the selected GPU runtime fails, existing CPU fallback may recover when enabled, but CPU has its own memory/runtime needs.

PlatformUse a compatible OS/architecture/runtime or another model. These estimates do not add hardware detection or block runs.

Benchmark evidence

Saved desktop measurements apply to the recorded hardware and runtime. Fallback cases report the device that actually ran; they do not prove GPU support on the requested device.

LateOn Code Base Desktop

4 configurations tested
Passed4
Failed0
TestedNov 14, 2026
Runtimetransformers 5.22.1

Quality criteria

Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.

PrecisionDeviceResultPlatformInference / load
FP16mpsAcceleratedPassedSlowmacos · aarch64282 ms / 2042 ms
FP32mpsAcceleratedPassedNot measuredmacos · aarch6489 ms / 2127 ms
BF16mpsAcceleratedPassedNot measuredmacos · aarch6499 ms / 2151 ms
FP32cpuPassedNot measuredmacos · aarch6444 ms / 1952 ms
Test setup & sample details
Quality gate
Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.
Tested
2026-11-14T12:51:10.763501+00:00
Environment
Apple M3 Max on macOS 27.2. Managed native or isolated Chrome module workers.
Runtime
transformers 5.22.1
Artifact revision
ace431824f35db231178fc602e33296784762a2e
Fixture
{
  "file": "model-samples/lateon-code-deduplicate-query.json",
  "size": 193,
  "sha1": "aa4a1de14636660ffe22aa8f370645d65e11cace",
  "sha256": "92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a",
  "url": "https://assets.localai.tools/samples/v1/files/92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a/lateon-code-deduplicate-query.json",
  "durationSeconds": 0
}
Rerun when
artifact revision; runtime version; retrieval fixture; tokenizer or projection semantics
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        7.260654165132987,
        6.410418759089929,
        4.511352200528618
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 43.79645799053833,
      "durationMs": 44.002957991324365,
      "embeddingsPerSecond": 22.83289667433922,
      "outputDevice": "cpu",
      "peakResidentMemoryBytes": 827359232,
      "loadMs": 1951.5656669973396,
      "totalMs": 1995.362124987878,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "cpu",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "ace431824f35db231178fc602e33296784762a2e",
      "acceleration": "cpu",
      "requestedDevice": "cpu",
      "requestedPrecision": "fp32",
      "provider": "PyTorch CPU",
      "placement": {
        "cpu/torch.float32": 135
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        7.260652283831964,
        6.410419379247145,
        4.511353234819863
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 89.34395801043138,
      "durationMs": 90.23150004213676,
      "embeddingsPerSecond": 11.192698670045989,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 441171968,
      "loadMs": 2126.886249985546,
      "totalMs": 2216.2302079959773,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "ace431824f35db231178fc602e33296784762a2e",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp32",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float32": 135
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        7.259892168588747,
        6.410219618267109,
        4.51089780226257
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 282.07033302169293,
      "durationMs": 282.8935830038972,
      "embeddingsPerSecond": 3.545215086207219,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 1265385472,
      "loadMs": 2042.3553750151768,
      "totalMs": 2324.4257080368698,
      "grade": "slow",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp16",
      "fallback": false,
      "artifactRevision": "ace431824f35db231178fc602e33296784762a2e",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float16": 135
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        7.271052113139462,
        6.400134029076571,
        4.503192895716307
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 98.80658297333866,
      "durationMs": 99.77108403109014,
      "embeddingsPerSecond": 10.120783149335644,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 1262764032,
      "loadMs": 2151.114584004972,
      "totalMs": 2249.921166978311,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "bf16",
      "fallback": false,
      "artifactRevision": "ace431824f35db231178fc602e33296784762a2e",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "bf16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.bfloat16": 135
      },
      "cpuThreads": 4
    }
  }
]

LateOn Code Base Pretrain Desktop

1 configurations tested
Passed1
Failed0
TestedNov 14, 2026
Runtimeonnxruntime 1.24.4

Quality criteria

Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.

PrecisionDeviceResultPlatformInference / load
INT8cpuPassedNot measuredmacos · aarch649 ms / 1019 ms
Test setup & sample details
Quality gate
Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.
Tested
2026-11-14T12:59:19.422792+00:00
Environment
Apple M3 Max on macOS 27.2. Managed native or isolated Chrome module workers.
Runtime
onnxruntime 1.24.4
Artifact revision
71251a6ee61eee488de7e3ae29f5fb4c3c94699b
Fixture
{
  "file": "model-samples/lateon-code-deduplicate-query.json",
  "size": 193,
  "sha1": "aa4a1de14636660ffe22aa8f370645d65e11cace",
  "sha256": "92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a",
  "url": "https://assets.localai.tools/samples/v1/files/92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a/lateon-code-deduplicate-query.json",
  "durationSeconds": 0
}
Rerun when
artifact revision; runtime version; retrieval fixture; tokenizer or projection semantics
Per-case quality & failure details
[
  {
    "precision": "int8",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        10.275207624103347,
        10.181887254510418,
        9.55230181269835
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 8.54358304059133,
      "durationMs": 8.79870803328231,
      "embeddingsPerSecond": 117.04691055835828,
      "outputDevice": "cpu",
      "peakResidentMemoryBytes": 466354176,
      "loadMs": 1019.1666249884292,
      "totalMs": 1027.7102080290206,
      "grade": "interactive",
      "engine": "onnxruntime",
      "version": "1.24.4",
      "device": "cpu",
      "precision": "int8",
      "fallback": false,
      "artifactRevision": "71251a6ee61eee488de7e3ae29f5fb4c3c94699b",
      "acceleration": "cpu",
      "requestedDevice": "cpu",
      "requestedPrecision": "int8",
      "provider": "CPUExecutionProvider",
      "placement": {
        "provider": "CPUExecutionProvider"
      },
      "cpuThreads": 4
    }
  }
]

LateOn Code Edge 17M Desktop

4 configurations tested
Passed4
Failed0
TestedNov 14, 2026
Runtimetransformers 5.22.1

Quality criteria

Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.

PrecisionDeviceResultPlatformInference / load
FP32mpsAcceleratedPassedNot measuredmacos · aarch6483 ms / 2250 ms
FP16mpsAcceleratedPassedNot measuredmacos · aarch64107 ms / 2071 ms
BF16mpsAcceleratedPassedNot measuredmacos · aarch6474 ms / 2134 ms
FP32cpuPassedNot measuredmacos · aarch644 ms / 1918 ms
Test setup & sample details
Quality gate
Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.
Tested
2026-11-14T12:50:55.224458+00:00
Environment
Apple M3 Max on macOS 27.2. Managed native or isolated Chrome module workers.
Runtime
transformers 5.22.1
Artifact revision
4bcdf5ed93f791259eb130b577a240f753d68dd8
Fixture
{
  "file": "model-samples/lateon-code-deduplicate-query.json",
  "size": 193,
  "sha1": "aa4a1de14636660ffe22aa8f370645d65e11cace",
  "sha256": "92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a",
  "url": "https://assets.localai.tools/samples/v1/files/92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a/lateon-code-deduplicate-query.json",
  "durationSeconds": 0
}
Rerun when
artifact revision; runtime version; retrieval fixture; tokenizer or projection semantics
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        8.123682441775719,
        7.640633358296425,
        6.967428729622705
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 3.900459036231041,
      "durationMs": 4.120708035770804,
      "embeddingsPerSecond": 256.3800800652136,
      "outputDevice": "cpu",
      "peakResidentMemoryBytes": 399065088,
      "loadMs": 1918.1354999891482,
      "totalMs": 1922.0359590253793,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "cpu",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "4bcdf5ed93f791259eb130b577a240f753d68dd8",
      "acceleration": "cpu",
      "requestedDevice": "cpu",
      "requestedPrecision": "fp32",
      "provider": "PyTorch CPU",
      "placement": {
        "cpu/torch.float32": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        8.123682857754075,
        7.640632599956996,
        6.967429650689402
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 83.42308300780132,
      "durationMs": 84.40754195908085,
      "embeddingsPerSecond": 11.987089950949006,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 438239232,
      "loadMs": 2249.9058750108816,
      "totalMs": 2333.328958018683,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "4bcdf5ed93f791259eb130b577a240f753d68dd8",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp32",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float32": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        8.122411548118123,
        7.640043453569127,
        6.967112810219427
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 106.54995799995959,
      "durationMs": 107.32104198541492,
      "embeddingsPerSecond": 9.385268833239515,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 475447296,
      "loadMs": 2070.97908295691,
      "totalMs": 2177.5290409568697,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp16",
      "fallback": false,
      "artifactRevision": "4bcdf5ed93f791259eb130b577a240f753d68dd8",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float16": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        8.12141500389203,
        7.646662239569483,
        6.962840451426261
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 74.07458400120959,
      "durationMs": 74.73500003106892,
      "embeddingsPerSecond": 13.49990706641931,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 476512256,
      "loadMs": 2133.698541962076,
      "totalMs": 2207.7731259632856,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "bf16",
      "fallback": false,
      "artifactRevision": "4bcdf5ed93f791259eb130b577a240f753d68dd8",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "bf16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.bfloat16": 46
      },
      "cpuThreads": 4
    }
  }
]

LateOn Code Edge 17M Pretrain Desktop

4 configurations tested
Passed4
Failed0
TestedNov 14, 2026
Runtimetransformers 5.22.1

Quality criteria

Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.

PrecisionDeviceResultPlatformInference / load
FP32mpsAcceleratedPassedNot measuredmacos · aarch6472 ms / 2145 ms
FP16mpsAcceleratedPassedNot measuredmacos · aarch6480 ms / 2119 ms
BF16mpsAcceleratedPassedNot measuredmacos · aarch6479 ms / 2035 ms
FP32cpuPassedNot measuredmacos · aarch643 ms / 2086 ms
Test setup & sample details
Quality gate
Three authored code-ranking fixtures, exact token IDs and agreement with original native FP32 token vectors on browser.
Tested
2026-11-14T12:51:26.332996+00:00
Environment
Apple M3 Max on macOS 27.2. Managed native or isolated Chrome module workers.
Runtime
transformers 5.22.1
Artifact revision
4ca3a44b3093e72d48461aa6a67cfd5c0025c007
Fixture
{
  "file": "model-samples/lateon-code-deduplicate-query.json",
  "size": 193,
  "sha1": "aa4a1de14636660ffe22aa8f370645d65e11cace",
  "sha256": "92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a",
  "url": "https://assets.localai.tools/samples/v1/files/92f8a217d38938179b0d9c1f40bab01fbdd3aba3d593558f706a7d4714205e8a/lateon-code-deduplicate-query.json",
  "durationSeconds": 0
}
Rerun when
artifact revision; runtime version; retrieval fixture; tokenizer or projection semantics
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        11.008074937466475,
        10.778504563194172,
        10.575960693814473
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 3.3625829964876175,
      "durationMs": 3.610290994402021,
      "embeddingsPerSecond": 297.39042903760264,
      "outputDevice": "cpu",
      "peakResidentMemoryBytes": 400801792,
      "loadMs": 2086.4617089973763,
      "totalMs": 2089.824291993864,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "cpu",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "4ca3a44b3093e72d48461aa6a67cfd5c0025c007",
      "acceleration": "cpu",
      "requestedDevice": "cpu",
      "requestedPrecision": "fp32",
      "provider": "PyTorch CPU",
      "placement": {
        "cpu/torch.float32": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        11.008073825715142,
        10.778503737443472,
        10.575959315576611
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 72.18625000678003,
      "durationMs": 73.01433302927762,
      "embeddingsPerSecond": 13.853053731231029,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 444792832,
      "loadMs": 2145.1021659886464,
      "totalMs": 2217.2884159954265,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp32",
      "fallback": false,
      "artifactRevision": "4ca3a44b3093e72d48461aa6a67cfd5c0025c007",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp32",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float32": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        11.006673701443722,
        10.776653626608478,
        10.573959346602939
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 79.6476670075208,
      "durationMs": 80.51179099129513,
      "embeddingsPerSecond": 12.555295560704549,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 476250112,
      "loadMs": 2119.2244589910842,
      "totalMs": 2198.872125998605,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "fp16",
      "fallback": false,
      "artifactRevision": "4ca3a44b3093e72d48461aa6a67cfd5c0025c007",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "fp16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.float16": 46
      },
      "cpuThreads": 4
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "scores": [
        11.002703419042156,
        10.77625338234631,
        10.571748466557949
      ],
      "winner": 0,
      "expected": 0,
      "grade": "retrieval-pass"
    },
    "runtime": {
      "inferenceMs": 79.11824999609962,
      "durationMs": 80.26395802153274,
      "embeddingsPerSecond": 12.639308883213394,
      "outputDevice": "mps",
      "peakResidentMemoryBytes": 475267072,
      "loadMs": 2035.3750829817727,
      "totalMs": 2114.4933329778723,
      "grade": "interactive",
      "engine": "transformers",
      "version": "5.22.1",
      "device": "mps",
      "precision": "bf16",
      "fallback": false,
      "artifactRevision": "4ca3a44b3093e72d48461aa6a67cfd5c0025c007",
      "acceleration": "partial",
      "requestedDevice": "mps",
      "requestedPrecision": "bf16",
      "provider": "Metal/MPS",
      "placement": {
        "mps:0/torch.bfloat16": 46
      },
      "cpuThreads": 4
    }
  }
]
Release validation still required
  • Public model backup and dependent recovery are deferred.
  • Windows benchmarking is deferred.

Something not quite right?

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

Attachments Optional

Up to 3 files · 50 MB each

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