LibriCount
Models estimate how many speakers appear in a recording with overlapping voices.
The mixtures contain up to ten speakers.
- Made by
- Fabian-Robert Stöter and colleagues
- Released
- December 2017
- Kind
- Speech recognition and translation
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Prediction
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Generated
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Audio and speech
- Last model added
- November 2023
- Scores refreshed
- November 20th, 2026
openai/whisper-medium tops the leaderboard with 57.9% Whisper large-v3 leads with 57.2%
Best score per entry from MTEB’s 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 Whisper large-v3, at 57.2%. This view shows only models we track, each at its best.
- 57.9 openai/whisper-medium
- 57.2 Whisper large-v3
- 55.0 facebook/wav2vec2-base
- 53.4 openai/whisper-small
- 52.3 microsoft/wavlm-large
- 52.1 microsoft/wavlm-base-plus
- 52.1 microsoft/wavlm-base-plus-sd
- 52.1 microsoft/wavlm-base-plus-sv
- 50.7 microsoft/wavlm-base
- 50.7 microsoft/wavlm-base-sd
- 50.7 microsoft/wavlm-base-sv
- 50.2 facebook/wav2vec2-xls-r-2b
- 50.2 openai/whisper-base
- 50.2 openai/whisper-tiny
- 49.8 facebook/wav2vec2-xls-r-1b
- 49.6 Qwen/Qwen2-Audio-7B
- 48.9 facebook/hubert-base-ls960
- 48.7 laion/clap-htsat-unfused
- 48.5 laion/larger_clap_general
- 48.4 facebook/data2vec-audio-base-960h
- 48.0 laion/larger_clap_music_and_speech
- 47.8 laion/clap-htsat-fused
- 46.1 facebook/wav2vec2-base-960h
- 45.6 google/vggish
- 45.4 microsoft/unispeech-sat-base-100h-libri-ft
- 45.3 facebook/wav2vec2-lv-60-espeak-cv-ft
- 45.2 facebook/wav2vec2-large
- 44.3 facebook/hubert-large-ls960-ft
- 44.1 facebook/seamless-m4t-v2-large
- 43.4 asapp/sew-d-tiny-100k-ft-ls100h
- 43.3 microsoft/speecht5_multimodal
- 43.1 asapp/sew-d-mid-400k-ft-ls100h
- 42.7 asapp/sew-d-base-plus-400k-ft-ls100h
- 42.5 vitouphy/wav2vec2-xls-r-300m-phoneme
- 42.5 facebook/data2vec-audio-large-960h
- 42.1 MIT/ast-finetuned-audioset-10-10-0.4593
- 42.1 microsoft/msclap-2023
- 41.8 facebook/mms-1b-all
- 41.3 google/yamnet
- 41.2 facebook/mms-1b-l1107
- 39.7 microsoft/msclap-2022
- 39.6 facebook/mms-1b-fl102
- 37.5 facebook/wav2vec2-xls-r-300m
- 36.7 OpenMuQ/MuQ-MuLan-large
- 34.9 speechbrain/m-ctc-t-large
- 34.6 lyrebird/wav2clip
- 33.2 LCO-Embedding/LCO-Embedding-Omni-7B
- 30.4 LCO-Embedding/LCO-Embedding-Omni-3B
- 28.4 speechbrain/cnn14-esc50
- 25.1 facebook/wav2vec2-large-xlsr-53
- 21.2 facebook/encodec_24khz
- 20.5 laion/larger_clap_music
- 1 openai/whisper-medium 57.9%
- 2 Whisper large-v3 57.2%
- 3 facebook/wav2vec2-base 55%
- 4 openai/whisper-small 53.4%
- 5 microsoft/wavlm-large 52.3%
- 6 microsoft/wavlm-base-plus 52.1%
- 7 microsoft/wavlm-base-plus-sd 52.1%
- 8 microsoft/wavlm-base-plus-sv 52.1%
- 9 microsoft/wavlm-base 50.7%
- 10 microsoft/wavlm-base-sd 50.7%
- 11 microsoft/wavlm-base-sv 50.7%
- 12 facebook/wav2vec2-xls-r-2b 50.2%
- 13 openai/whisper-base 50.2%
- 14 openai/whisper-tiny 50.2%
- 15 facebook/wav2vec2-xls-r-1b 49.8%
- 16 Qwen/Qwen2-Audio-7B 49.6%
- 17 facebook/hubert-base-ls960 48.9%
- 18 laion/clap-htsat-unfused 48.7%
- 19 laion/larger_clap_general 48.5%
- 20 facebook/data2vec-audio-base-960h 48.4%
- 21 laion/larger_clap_music_and_speech 48%
- 22 laion/clap-htsat-fused 47.8%
- 23 facebook/wav2vec2-base-960h 46.1%
- 24 google/vggish 45.6%
- 25 microsoft/unispeech-sat-base-100h-libri-ft 45.4%
- 26 facebook/wav2vec2-lv-60-espeak-cv-ft 45.3%
- 27 facebook/wav2vec2-large 45.2%
- 28 facebook/hubert-large-ls960-ft 44.3%
- 29 facebook/seamless-m4t-v2-large 44.1%
- 30 asapp/sew-d-tiny-100k-ft-ls100h 43.4%
- 31 microsoft/speecht5_multimodal 43.3%
- 32 asapp/sew-d-mid-400k-ft-ls100h 43.1%
- 33 asapp/sew-d-base-plus-400k-ft-ls100h 42.7%
- 34 vitouphy/wav2vec2-xls-r-300m-phoneme 42.5%
- 35 facebook/data2vec-audio-large-960h 42.5%
- 36 MIT/ast-finetuned-audioset-10-10-0.4593 42.1%
- 37 microsoft/msclap-2023 42.1%
- 38 facebook/mms-1b-all 41.8%
- 39 google/yamnet 41.3%
- 40 facebook/mms-1b-l1107 41.2%
- 41 microsoft/msclap-2022 39.7%
- 42 facebook/mms-1b-fl102 39.6%
- 43 facebook/wav2vec2-xls-r-300m 37.5%
- 44 OpenMuQ/MuQ-MuLan-large 36.7%
- 45 speechbrain/m-ctc-t-large 34.9%
- 46 lyrebird/wav2clip 34.6%
- 47 LCO-Embedding/LCO-Embedding-Omni-7B 33.2%
- 48 LCO-Embedding/LCO-Embedding-Omni-3B 30.4%
- 49 speechbrain/cnn14-esc50 28.4%
- 50 facebook/wav2vec2-large-xlsr-53 25.1%
- 51 facebook/encodec_24khz 21.2%
- 52 laion/larger_clap_music 20.5%
Source: MTEB’s leaderboard
Source: MTEB’s leaderboard
| # | Entry | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | openai/whisper-medium open weights, 0.8B | 57.9% | huggingface.co | Official leaderboard | ||
| 2 | Whisper large-v3 | 57.2% | Nov 2023 | huggingface.co | Official leaderboard | |
| 3 | facebook/wav2vec2-base open weights, 0.1B | 55% | huggingface.co | Official leaderboard | ||
| 4 | openai/whisper-small open weights, 0.2B | 53.4% | huggingface.co | Official leaderboard | ||
| 5 | microsoft/wavlm-large open weights, 0.3B | 52.3% | huggingface.co | Official leaderboard | ||
| 6 | microsoft/wavlm-base-plus open weights, 0.1B | 52.1% | huggingface.co | Official leaderboard | ||
| 7 | microsoft/wavlm-base-plus-sd open weights, 0.1B | 52.1% | huggingface.co | Official leaderboard | ||
| 8 | microsoft/wavlm-base-plus-sv open weights, 0.1B | 52.1% | huggingface.co | Official leaderboard | ||
| 9 | microsoft/wavlm-base open weights, 0.1B | 50.7% | huggingface.co | Official leaderboard | ||
| 10 | microsoft/wavlm-base-sd open weights, 0.1B | 50.7% | huggingface.co | Official leaderboard | ||
| 11 | microsoft/wavlm-base-sv open weights, 0.1B | 50.7% | huggingface.co | Official leaderboard | ||
| 12 | facebook/wav2vec2-xls-r-2b open weights, 2B | 50.2% | huggingface.co | Official leaderboard | ||
| 13 | openai/whisper-base open weights, 0.1B | 50.2% | huggingface.co | Official leaderboard | ||
| 14 | openai/whisper-tiny open weights, 0B | 50.2% | huggingface.co | Official leaderboard | ||
| 15 | facebook/wav2vec2-xls-r-1b open weights, 1B | 49.8% | huggingface.co | Official leaderboard | ||
| 16 | Qwen/Qwen2-Audio-7B open weights, 7B | 49.6% | huggingface.co | Official leaderboard | ||
| 17 | facebook/hubert-base-ls960 open weights, 0.1B | 48.9% | huggingface.co | Official leaderboard | ||
| 18 | laion/clap-htsat-unfused open weights, 0.2B | 48.7% | huggingface.co | Official leaderboard | ||
| 19 | laion/larger_clap_general open weights, 0.2B | 48.5% | huggingface.co | Official leaderboard | ||
| 20 | facebook/data2vec-audio-base-960h open weights, 0.1B | 48.4% | huggingface.co | Official leaderboard | ||
| 21 | laion/larger_clap_music_and_speech open weights, 0.2B | 48% | huggingface.co | Official leaderboard | ||
| 22 | laion/clap-htsat-fused open weights, 0.2B | 47.8% | huggingface.co | Official leaderboard | ||
| 23 | facebook/wav2vec2-base-960h open weights, 0.1B | 46.1% | huggingface.co | Official leaderboard | ||
| 24 | google/vggish open weights, 0.1B | 45.6% | github.com | Official leaderboard | ||
| 25 | microsoft/unispeech-sat-base-100h-libri-ft open weights, 0.1B | 45.4% | huggingface.co | Official leaderboard | ||
| 26 | facebook/wav2vec2-lv-60-espeak-cv-ft open weights, 0.3B | 45.3% | huggingface.co | Official leaderboard | ||
| 27 | facebook/wav2vec2-large open weights, 0.3B | 45.2% | huggingface.co | Official leaderboard | ||
| 28 | facebook/hubert-large-ls960-ft open weights, 0.3B | 44.3% | huggingface.co | Official leaderboard | ||
| 29 | facebook/seamless-m4t-v2-large open weights, 2.3B | 44.1% | huggingface.co | Official leaderboard | ||
| 30 | asapp/sew-d-tiny-100k-ft-ls100h open weights, 0B | 43.4% | huggingface.co | Official leaderboard | ||
| 31 | microsoft/speecht5_multimodal open weights, 0.3B | 43.3% | huggingface.co | Official leaderboard | ||
| 32 | asapp/sew-d-mid-400k-ft-ls100h open weights, 0.1B | 43.1% | huggingface.co | Official leaderboard | ||
| 33 | asapp/sew-d-base-plus-400k-ft-ls100h open weights, 0.1B | 42.7% | huggingface.co | Official leaderboard | ||
| 34 | vitouphy/wav2vec2-xls-r-300m-phoneme open weights, 0.3B | 42.5% | huggingface.co | Official leaderboard | ||
| 35 | facebook/data2vec-audio-large-960h open weights, 0.3B | 42.5% | huggingface.co | Official leaderboard | ||
| 36 | MIT/ast-finetuned-audioset-10-10-0.4593 open weights, 0.1B | 42.1% | huggingface.co | Official leaderboard | ||
| 37 | microsoft/msclap-2023 open weights, 0.2B | 42.1% | github.com | Official leaderboard | ||
| 38 | facebook/mms-1b-all open weights, 1B | 41.8% | huggingface.co | Official leaderboard | ||
| 39 | google/yamnet open weights, 0B | 41.3% | tfhub.dev | Official leaderboard | ||
| 40 | facebook/mms-1b-l1107 open weights, 1B | 41.2% | huggingface.co | Official leaderboard | ||
| 41 | microsoft/msclap-2022 open weights, 0.2B | 39.7% | github.com | Official leaderboard | ||
| 42 | facebook/mms-1b-fl102 open weights, 1B | 39.6% | huggingface.co | Official leaderboard | ||
| 43 | facebook/wav2vec2-xls-r-300m open weights, 0.3B | 37.5% | huggingface.co | Official leaderboard | ||
| 44 | OpenMuQ/MuQ-MuLan-large open weights, 0.6B | 36.7% | huggingface.co | Official leaderboard | ||
| 45 | speechbrain/m-ctc-t-large open weights, 1.1B | 34.9% | huggingface.co | Official leaderboard | ||
| 46 | lyrebird/wav2clip open weights, 0.2B | 34.6% | github.com | Official leaderboard | ||
| 47 | LCO-Embedding/LCO-Embedding-Omni-7B open weights, 8.9B | 33.2% | huggingface.co | Official leaderboard | ||
| 48 | LCO-Embedding/LCO-Embedding-Omni-3B open weights, 4.7B | 30.4% | huggingface.co | Official leaderboard | ||
| 49 | speechbrain/cnn14-esc50 open weights, 0.1B | 28.4% | huggingface.co | Official leaderboard | ||
| 50 | facebook/wav2vec2-large-xlsr-53 open weights, 0.3B | 25.1% | huggingface.co | Official leaderboard | ||
| 51 | facebook/encodec_24khz open weights, 0B | 21.2% | huggingface.co | Official leaderboard | ||
| 52 | laion/larger_clap_music open weights, 0.2B | 20.5% | huggingface.co | Official leaderboard |
| # | Model | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | Whisper large-v3 | 57.2% | Nov 2023 | huggingface.co | Official leaderboard |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: MTEB’s 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 3 years ago, in November 2023, for Whisper large-v3. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (52 entries).
- Pass Models scored 1 model we track, plus 51 other entries on its leaderboard.
- Warning Newest score The newest came 3 years ago, in November 2023, for Whisper large-v3.
- Note Frontier models None of its models are on the capabilities index.
- Pass Headroom The best is 57.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on MTEB: 1 row for models we track and 51 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.