LibriSpeech
Voxtral Small 1.53% leads since July 2025 Models transcribe English audiobook speech. Scored by word error rate on the test-clean and test-other splits.
About 1,000 hours of read speech from LibriVox. Released at ICASSP 2015.
- Released
- April 2015
- Kind
- Speech recognition and translation
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Last in a launch post
- April 2026, dropped before it saturated
- Last model added
- April 2026
Voxtral Small leads with 1.53% among models released by January 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in January 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 Voxtral Small 1.53%
- 2
Qwen3-ASR-1.7B 1.63% - 3 MiMo-Audio-7B-Instruct 3.76%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Voxtral Small | 1.53% | Jul 2025 | Lab model card | |
| 2 | Qwen3-ASR-1.7B | 1.63% | Jan 2026 | Lab model card | |
| 3 | MiMo-Audio-7B-Instruct | 3.76% | Sep 2025 | Lab model card |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs headline it less often now, last in April 2026
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Apr 28, 2026 |  Seed2.0 Lite (0428) | 1.62 | Mean of Librispeech test-clean 1.07 / test-other 2.17 (WER, lower is better) |
| Mar 30, 2026 | Qwen3.5-Omni-Plus | 1.67 | Mean of Librispeech test-clean 1.11 / test-other 2.23 |
| Mar 26, 2026 |  Cohere Transcribe | 1.25 | LS clean (LibriSpeech test-clean) WER |
| Jan 29, 2026 | Qwen3-ASR-1.7B | 1.63 | Librispeech clean WER, Qwen3-ASR-1.7B |
| Oct 31, 2025 |  LongCat-Flash-Omni | 2.79 | Mean of LibriSpeech test-clean 1.57 / test-other 4.01 |
| Sep 22, 2025 | Qwen3-Omni-30B-A3B | 1.85 | Mean of Librispeech test-clean 1.22 / test-other 2.48 |
| Apr 25, 2025 | Kimi-Audio-7B-Instruct | 1.85 | Mean of LibriSpeech test-clean 1.28 / test-other 2.42 |
| Mar 27, 2025 | Qwen2.5-Omni-7B | 2.6 | Mean of Librispeech test-clean 1.8 / test-other 3.4 (dev-clean 1.6, dev-other 3.5 not used) |
8 system cards report it, from 6 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Apr 17, 2026 | Qwen3.5-Omni-Plus technical report | Librispeech ASR WER test-clean 1.11, test-other 2.23, no overall (Table 5) | Capabilities | |
| Feb 4, 2026 |  ERNIE 5.0 technical report | LibriSpeech clean 1.16, other 2.61, WER, lower is better, no overall score (Table 7) | Capabilities | |
| Oct 31, 2025 | LongCat-Flash-Omni technical report | LibriSpeech test-clean 1.57, test-other 4.01, WER, lower is better, no overall score (Table 9) | Capabilities | |
| Sep 22, 2025 | Qwen3-Omni-30B-A3B technical report | Librispeech clean 1.22, other 2.48, WER, lower is better, Instruct, no overall score (Table 6) | Capabilities | |
| Sep 18, 2025 | MiMo-Audio-7B-Instruct technical report | 3.76 | Librispeech-test-clean WER, lower is better (Table 9) | Capabilities |
| Jul 17, 2025 | Voxtral Small technical report | 1.53 | LibriSpeech Test Clean WER, lower is better, Test Other 3.14 (Table 3) | Capabilities |
| Apr 25, 2025 | Kimi-Audio-7B-Instruct technical report | LibriSpeech test-clean 1.28, test-other 2.42, WER, lower is better, no overall score (Table 4) | Capabilities | |
| Mar 26, 2025 | Qwen2.5-Omni-7B technical report | Librispeech dev-clean 1.6, dev-other 3.5, test-clean 1.8, test-other 3.4, WER, lower is better, no overall score (Table 2) | Capabilities |
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%