Voice Arena Diarization Bench
Models identify who spoke when in recorded conversations. The board measures diarization error and publishes a separate track for speaker attributed transcription.
The release reports speaker confusion, missed speech, and false alarms.
- Made by
- Voice Arena
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Measures
- Capability
- In and out
- Audio → Time series, Text
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- A host
- In views
- Audio and speech
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
Nemotron 3 Diarization tops the leaderboard with 4.29% Muse Voice Transcribe 1.0 leads with 31%
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Muse Voice Transcribe 1.0, at 31%. This view shows only models we track, each at its best.
- 1 Nemotron 3 Diarization Diarization only / 250 ms collar / Standard / offline 4.29%
- 2 Precision-3 Diarization only / 250 ms collar / Standard / offline 9.43%
- 3 DiariZen WavLM Large s80 MD v2 Diarization only / 250 ms collar / Standard / offline 9.83%
- 4 Pulse With ASR / 250 ms collar / Standard / realtime 12.4%
- 5 Precision-2 Diarization only / 250 ms collar / Standard / offline 13.4%
- 6 Streaming Sortformer Diarizer 4spk v2 Diarization only / 250 ms collar / Standard / realtime 14.3%
- 7 Community-1 Diarization only / 250 ms collar / Standard / offline 19.9%
- 8 Voxtral Mini Transcribe 3 With ASR / 250 ms collar / Standard / offline 30.5%
- 9
Muse Voice Transcribe 1.0 With ASR / 250 ms collar / Standard / realtime 31%
- 10 Voxtral Small Transcribe 3 With ASR / 250 ms collar / Standard / offline 31.2%
- 11 Scribe v2 With ASR / 250 ms collar / Standard / offline 31.9%
- 12 Universal-3.5 Pro With ASR / 250 ms collar / Word-level, range hint 2 to 8 / realtime 34.4%
- 13 Soniox v5 Async With ASR / 250 ms collar / Standard / offline 41%
- 14 Enhanced With ASR / 250 ms collar / Standard / realtime 44%
- 15 Nova-3 With ASR / 250 ms collar / Word-level / realtime 59.9%
Source: Official Voice Arena Diarization Bench leaderboard
Source: Official Voice Arena Diarization Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Nemotron 3 Diarization open weights | 4.29% | Diarization only / 250 ms collar / Standard / offline | Official leaderboard | |
| 2 | Precision-3 closed weights | 9.43% | Diarization only / 250 ms collar / Standard / offline | Official leaderboard | |
| 3 | DiariZen WavLM Large s80 MD v2 open weights | 9.83% | Diarization only / 250 ms collar / Standard / offline | Official leaderboard | |
| 4 | Pulse closed weights | 12.4% | With ASR / 250 ms collar / Standard / realtime | Official leaderboard | |
| 5 | Precision-2 closed weights | 13.4% | Diarization only / 250 ms collar / Standard / offline | Official leaderboard | |
| 6 | Streaming Sortformer Diarizer 4spk v2 open weights | 14.3% | Diarization only / 250 ms collar / Standard / realtime | Official leaderboard | |
| 7 | Community-1 open weights | 19.9% | Diarization only / 250 ms collar / Standard / offline | Official leaderboard | |
| 8 | Voxtral Mini Transcribe 3 closed weights | 30.5% | With ASR / 250 ms collar / Standard / offline | Official leaderboard | |
| 9 | | 31% | With ASR / 250 ms collar / Standard / realtime | Sep 2026 | Official leaderboard |
| 10 | Voxtral Small Transcribe 3 closed weights | 31.2% | With ASR / 250 ms collar / Standard / offline | Official leaderboard | |
| 11 | Scribe v2 closed weights | 31.9% | With ASR / 250 ms collar / Standard / offline | Official leaderboard | |
| 12 | Universal-3.5 Pro closed weights | 34.4% | With ASR / 250 ms collar / Word-level, range hint 2 to 8 / realtime | Official leaderboard | |
| 13 | Soniox v5 Async closed weights | 41% | With ASR / 250 ms collar / Standard / offline | Official leaderboard | |
| 14 | Enhanced closed weights | 44% | With ASR / 250 ms collar / Standard / realtime | Official leaderboard | |
| 15 | Nova-3 closed weights | 59.9% | With ASR / 250 ms collar / Word-level / realtime | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 31% | With ASR / 250 ms collar / Standard / realtime | Sep 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Voice Arena Diarization Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Voice Arena Diarization Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (15 entries).
- Pass Models scored 1 model we track, plus 14 other entries on its leaderboard.
- Pass Newest score The newest is from September 2026, for Muse Voice Transcribe 1.0.
- Note Frontier models None of its models are on the capabilities index.
- Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
- Pass Importer We read the board on voicearena.com: 3 rows for models we track and 14 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.