Clotho
Models write a caption describing what can be heard in an audio recording.
MTEB also uses the paired recordings and captions to evaluate cross-modal retrieval.
- Made by
- Konstantinos Drossos and colleagues
- Released
- October 2019
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Audio and speech
EnCLAP++ tops the leaderboard with 0.48
Best score per entry from SOTA2 Research’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.
- 0.48 EnCLAP++
- 0.46 EnCLAP-large
- 0.46 EnCLAP-base
- 0.44 Qwen-Audio
- 0.43 M2D-CLAP2025 stage2
- 0.43 M2D-CLAP2025 stage2.1
- 0.43 M2D-CLAP2025
- 0.41 M2D-CLAP2025 stage1
- 0.39 Prefix Tuning for Automated Audio Captioning
- 0.39 DiDi-Merging
- 0.33 MiCo-Chat-7B
- 0.18 LTBM (local)
- 0.18 RNN-GRU-EncDec + VGGish + Word2Vec
- 0.15 RNN-GRU-EncDec + VGGish
- 0.11 RNN-GRU-EncDec + Log Mel Energy
- 1 EnCLAP++ 0.48
- 2 EnCLAP-large training setup: Trained on AudioCaps and Clotho 0.46
- 3 EnCLAP-base training setup: Trained on AudioCaps and Clotho 0.46
- 4 Qwen-Audio zero-shot: true 0.44
- 5 M2D-CLAP2025 stage2 Audio encoder AudioSet fine-tuning: false 0.43
- 6 M2D-CLAP2025 stage2.1 Audio encoder AudioSet fine-tuning: false 0.43
- 7 M2D-CLAP2025 Audio encoder AudioSet fine-tuning: true 0.43
- 8 M2D-CLAP2025 stage1 Audio encoder AudioSet fine-tuning: false 0.41
- 9 Prefix Tuning for Automated Audio Captioning Training set: Clotho 0.39
- 10 DiDi-Merging 0.39
- 11 MiCo-Chat-7B 0-shot: true 0.33
- 12 LTBM (local) Keep Ratio: 0.75 0.18
- 13 RNN-GRU-EncDec + VGGish + Word2Vec Input Features: VGGish, Word Embeddings: Word2Vec 0.18
- 14 RNN-GRU-EncDec + VGGish Input Features: VGGish 0.15
- 15 RNN-GRU-EncDec + Log Mel Energy Input Features: Log Mel Energy 0.11
Source: SOTA2 Research’s leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | EnCLAP++ | 0.48 | Official leaderboard | ||
| 2 | EnCLAP-large | 0.46 | training setup: Trained on AudioCaps and Clotho | Official leaderboard | |
| 3 | EnCLAP-base | 0.46 | training setup: Trained on AudioCaps and Clotho | Official leaderboard | |
| 4 | Qwen-Audio | 0.44 | zero-shot: true | Official leaderboard | |
| 5 | M2D-CLAP2025 stage2 | 0.43 | Audio encoder AudioSet fine-tuning: false | Official leaderboard | |
| 6 | M2D-CLAP2025 stage2.1 | 0.43 | Audio encoder AudioSet fine-tuning: false | Official leaderboard | |
| 7 | M2D-CLAP2025 | 0.43 | Audio encoder AudioSet fine-tuning: true | Official leaderboard | |
| 8 | M2D-CLAP2025 stage1 | 0.41 | Audio encoder AudioSet fine-tuning: false | Official leaderboard | |
| 9 | Prefix Tuning for Automated Audio Captioning | 0.39 | Training set: Clotho | Official leaderboard | |
| 10 | DiDi-Merging | 0.39 | Official leaderboard | ||
| 11 | MiCo-Chat-7B | 0.33 | 0-shot: true | Official leaderboard | |
| 12 | LTBM (local) | 0.18 | Keep Ratio: 0.75 | Official leaderboard | |
| 13 | RNN-GRU-EncDec + VGGish + Word2Vec | 0.18 | Input Features: VGGish, Word Embeddings: Word2Vec | Official leaderboard | |
| 14 | RNN-GRU-EncDec + VGGish | 0.15 | Input Features: VGGish | Official leaderboard | |
| 15 | RNN-GRU-EncDec + Log Mel Energy | 0.11 | Input Features: Log Mel Energy | Official 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 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the official leaderboard (15 entries).
- Pass Models scored 15 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on SOTA2 Research: 15 entries at the last read, none a model we track.
- 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.