AssemblyAI Speech Benchmarks
Speech recognition systems transcribe clean, accented, multilingual and code-switched audio, identify entities and diarize speakers. Error rates compare production systems across public speech sets.
AssemblyAI reports results for 15 current systems across word error rate, entity error, diarization, code switching and multilingual speech. The page covers Common Voice, FLEURS, VoxPopuli, DiPCo, CALLHOME, NOTSOFAR, AMI and related sets, with runs through November 14th, 2026.
- Made by
- AssemblyAI
- Released
- November 2026
- Kind
- Speech recognition and translation
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- Many languages
- Interaction
- Single turn
- Items from
- Another benchmark
- Refresh
- New items on a schedule
- Run by
- A host
- In views
- Audio and speech, Many languages, Kept fresh
- Last model added
- August 2026
- Scores refreshed
- November 21st, 2026
AssemblyAI-STT-3.7 tops the leaderboard with 2.10% Voxtral Mini Transcribe 2 leads with 5.60% among models released by August 2026
Best score per entry from the benchmark's own leaderboard, read on November 21st, 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 Voxtral Mini Transcribe 2, at 5.60%. This view shows only models we track, each at its best.
The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 AssemblyAI-STT-3.7 2.10%
- 2 AssemblyAI Universal-3.5 Pro (previous) 2.50%
- 3 ElevenLabs Scribe v2 4.10%
- 4
Voxtral Mini Transcribe 2 5.60%
- 5 AWS Transcribe 6.10%
- 6
Gemini 3.5 Transcribe 7.30%
- 7 Soniox stt-async-v5 7.40%
- 8 Cohere Transcribe (language pinned) 8.50%
- 9 Speechmatics Enhanced 8.80%
- 10 Azure Batch Transcription 9.70%
- 11 gpt-4o-transcribe 9.80%
- 12 Gladia Solaria-1 10.3%
- 13 Deepgram Nova-3 10.4%
- 14 Grok STT 10.5%
- 15 Parakeet TDT 0.6B v3 11.3%
Source: Official AssemblyAI Speech Benchmarks leaderboard
- 1
Voxtral Mini Transcribe 2 5.60%
- 2
Gemini 3.5 Transcribe 7.30%
- 3 gpt-4o-transcribe 9.80%
Source: Official AssemblyAI Speech Benchmarks leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | AssemblyAI-STT-3.7 | 2.10% | Official leaderboard | ||
| 2 | AssemblyAI Universal-3.5 Pro (previous) | 2.50% | Official leaderboard | ||
| 3 | ElevenLabs Scribe v2 | 4.10% | Official leaderboard | ||
| 4 | | 5.60% | Feb 2026 | Official leaderboard | |
| 5 | AWS Transcribe | 6.10% | Official leaderboard | ||
| 6 | | 7.30% | Aug 2026 | Official leaderboard | |
| 7 | Soniox stt-async-v5 | 7.40% | Official leaderboard | ||
| 8 | Cohere Transcribe (language pinned) | 8.50% | Official leaderboard | ||
| 9 | Speechmatics Enhanced | 8.80% | Official leaderboard | ||
| 10 | Azure Batch Transcription | 9.70% | Official leaderboard | ||
| 11 | gpt-4o-transcribe | 9.80% | Mar 2025 | Official leaderboard | |
| 12 | Gladia Solaria-1 | 10.3% | Official leaderboard | ||
| 13 | Deepgram Nova-3 | 10.4% | Official leaderboard | ||
| 14 | Grok STT | 10.5% | Official leaderboard | ||
| 15 | Parakeet TDT 0.6B v3 | 11.3% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 5.60% | Feb 2026 | Official leaderboard | |
| 2 | | 7.30% | Aug 2026 | Official leaderboard | |
| 3 | gpt-4o-transcribe | 9.80% | Mar 2025 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as WER, clean, WER, accented and Entities. Voxtral Mini Transcribe 2 leads 3 of them, and Gemini 3.5 Transcribe leads the other one. Arrows show how far a model moved from its overall rank among the models we track.
WER, clean
Mean word error rate on clean-reference speech. 3 models we track have a score here.
- 1
Voxtral Mini Transcribe 2 5.30%
- 2 gpt-4o-transcribe ↑1 5.50%
- 3
Gemini 3.5 Transcribe ↓1 7.00%
WER, accented
Mean word error rate on accented English speech. 3 models we track have a score here.
- 1
Voxtral Mini Transcribe 2 8.70%
- 2
Gemini 3.5 Transcribe 9.20%
- 3 gpt-4o-transcribe 11.8%
Entities
Mean missed-entity error rate. 3 models we track have a score here.
- 1
Gemini 3.5 Transcribe ↑1 14.5%
- 2
Voxtral Mini Transcribe 2 ↓1 21.5%
- 3 gpt-4o-transcribe 22.9%
Multilingual
Mean multilingual error rate. 3 models we track have a score here.
- 1
Voxtral Mini Transcribe 2 7.20%
- 2
Gemini 3.5 Transcribe 7.80%
- 3 gpt-4o-transcribe 8.70%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official AssemblyAI Speech Benchmarks 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 AssemblyAI Speech Benchmarks 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
All 9 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All 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 3 models we track, plus 12 other entries on its leaderboard.
- Pass Newest score The newest is from August 2026, for Gemini 3.5 Transcribe.
- 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 assemblyai.com: 3 rows for models we track and 12 other entries at the last read.
- Pass Last read November 21st, 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.