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.

Back to models

Voxtral Mini Realtime 2602 Desktop

Transcribe speech into text with incremental output while processing audio.

Desktop inferencePrivate · Runs on your computer
About this model

Transcribe speech into text with incremental output while processing audio.

Created by
Mistral
License
Apache-2.0
Model released
Feb 4, 2026
Source repository created
Jan 21, 2026
Model weights · Q4
2.8 GB

Best for

  • Local captions and transcription of recordings in spoken languages supported by the checkpoint.

Keep in mind

  • Background noise, overlapping speakers, names and accents can reduce accuracy. Verify important transcripts.
  • Output contains transcription text. Speaker identity and word timestamps are unavailable in this adapter.
  • Samples cover English and Spanish. Other upstream languages are not quality-certified here.

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

Q4 · metal (verified platform only)

Sources & technical details
Original model
https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602
Release verification
Verified
Release evidence
Publisher launch announcement names the open-weight realtime checkpoint.
Source updated
Mar 11, 2026
Native model files
https://huggingface.co/handy-computer/Voxtral-Mini-4B-Realtime-2602-gguf
Files updated
Oct 16, 2026
Artifact revision
7fd3d424fc558edca2c53808578b0541ef52a81a

Desktop app

Sample result

Voxtral Mini Realtime 2602 Desktop

Run Voxtral Mini Realtime 2602 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 SiliconMETAL / Q4TestedNot establishedRAM estimate
macOS Apple SiliconCPU / Q4TestedNot establishedRAM 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.

Voxtral Mini Realtime 2602 Desktop

6 configurations tested
Passed6
Failed0
TestedNov 14, 2026
Runtimetranscribe.cpp 0.4.2

Quality criteria

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

PrecisionDeviceResultPlatformInference / loadTask metric
Q4metalAcceleratedPassedFastmacos · aarch641431 ms / 337 ms0.246 RTF
Q4metalAcceleratedPassedFastmacos · aarch641213 ms / 337 ms0.269 RTF
Q4metalAcceleratedPassedFastmacos · aarch643123 ms / 337 ms0.284 RTF
Q4cpuPassedFastmacos · aarch6410191 ms / 299 ms0.926 RTF
Q4cpuPassedModeratemacos · aarch646238 ms / 299 ms1.075 RTF
Q4cpuPassedModeratemacos · aarch645295 ms / 299 ms1.175 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-14T20:44:15.678869+00:00
Environment
{ "os": "macos", "osVersion": "27.2", "arch": "aarch64", "hardware": "Apple M3 Max" }
Runtime
transcribe.cpp 0.4.2
Artifact revision
7fd3d424fc558edca2c53808578b0541ef52a81a
Fixture
{
  "file": "model-samples/desktop/voxtral-realtime/jfk.wav",
  "size": 352078,
  "sha256": "4d968ac99a1d0d4bc42ae8dd1552f4235e7b952a8121b39797f3dc2ca16123e9",
  "sha1": "069db705beb788608e93b99f1a4300451ebfb972",
  "url": "https://assets.localai.tools/samples/v1/files/4d968ac99a1d0d4bc42ae8dd1552f4235e7b952a8121b39797f3dc2ca16123e9/jfk.wav"
}
Rerun when
Artifact revision; Runtime version; Fixture checksum; Quality threshold
Per-case quality & failure details
[
  {
    "precision": "q4",
    "device": "metal",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "reference": "And so my fellow Americans ask not what your country can do for you ask what you can do for your country."
    },
    "runtime": {
      "inferenceMs": 3122.85870796768,
      "firstTokenMs": 1345.7481670193374,
      "audioSeconds": 11,
      "realTimeFactor": 0.28389624617888,
      "device": "metal",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 337.1142499963753,
      "totalMs": 3459.9729579640552,
      "grade": "fast",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  },
  {
    "precision": "q4",
    "device": "metal",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "reference": "The morning train arrives at nine. Please bring your ticket."
    },
    "runtime": {
      "inferenceMs": 1213.1156669929624,
      "firstTokenMs": 501.85270799556747,
      "audioSeconds": 4.5046875,
      "realTimeFactor": 0.2693007377299674,
      "device": "metal",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 337.1142499963753,
      "totalMs": 1550.2299169893377,
      "grade": "fast",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  },
  {
    "precision": "q4",
    "device": "metal",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.18181818181818182,
      "reference": "El tren llega a las nueve. Por favor, trae tu billete."
    },
    "runtime": {
      "inferenceMs": 1430.7763330289163,
      "firstTokenMs": 692.114333040081,
      "audioSeconds": 5.805,
      "realTimeFactor": 0.24647309785166516,
      "device": "metal",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 337.1142499963753,
      "totalMs": 1767.8905830252916,
      "grade": "fast",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  },
  {
    "precision": "q4",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "reference": "And so my fellow Americans ask not what your country can do for you ask what you can do for your country."
    },
    "runtime": {
      "inferenceMs": 10191.472999984398,
      "firstTokenMs": 3180.08458300028,
      "audioSeconds": 11,
      "realTimeFactor": 0.9264975454531271,
      "device": "cpu",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 298.8963749958202,
      "totalMs": 10490.369374980219,
      "grade": "fast",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  },
  {
    "precision": "q4",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "reference": "The morning train arrives at nine. Please bring your ticket."
    },
    "runtime": {
      "inferenceMs": 5294.609333039261,
      "firstTokenMs": 2573.031250038184,
      "audioSeconds": 4.5046875,
      "realTimeFactor": 1.1753555231165895,
      "device": "cpu",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 298.8963749958202,
      "totalMs": 5593.505708035082,
      "grade": "moderate",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  },
  {
    "precision": "q4",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.18181818181818182,
      "reference": "El tren llega a las nueve. Por favor, trae tu billete."
    },
    "runtime": {
      "inferenceMs": 6237.999667006079,
      "firstTokenMs": 3237.7433329820633,
      "audioSeconds": 5.805,
      "realTimeFactor": 1.0745908125764132,
      "device": "cpu",
      "precision": "q4",
      "version": "0.4.2",
      "loadMs": 298.8963749958202,
      "totalMs": 6536.896042001899,
      "grade": "moderate",
      "artifactRevision": "7fd3d424fc558edca2c53808578b0541ef52a81a"
    }
  }
]
Release validation still required
  • Packaged UI, public backup and separate forced model/runtime source recovery.

Something not quite right?

Tell us what happened with Voxtral Mini Realtime 2602 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.