Benchmark Almanac

CodecMOS-Accent

CosyVoice 2 4 tops its leaderboard
Not in model cards

Speech systems reconstruct recordings or clone a voice across ten English accents. Listeners score naturalness, speaker similarity, and accent similarity separately.

Seven teams used the ratings in VoiceMOS 2026 Track 3, which held out 600 clips. The original listening study included 4,000 clips.

Released
March 2026
Kind
Audio understanding and voice
Level
Everyday
Scored
Graded by people
Answers
Open-ended
Measures
Preference
Language
English
In and out
Audio, Text → Audio
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Audio and speech
24
Entries in its paper, 0 models we track

CosyVoice 2 tops the leaderboard with 4

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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.

Bars start at 1.

  1. 1 CosyVoice 2 4
  2. 2 OpenAudio s1 mini 4
  3. 3 Llasa-1B 4
  4. 4 Chatterbox 4
  5. 5 NanoCodec 4
  6. 6 FACodec 4
  7. 7 NeuTTS Air 4
  8. 8 NeuCodec 4
  9. 9 IndexTTS2 4
  10. 10 TorToiSe 4
  11. 11 VoiceStar 4
  12. 12 XTTS 4
  13. 13 MaskGCT 4
  14. 14 VevoTTS 4
  15. 15 FireRedTTS 4
  16. 16 SNAC 3
  17. 17 Mimi 4.4kbps 3
  18. 18 WavTokenizer 40hz 3
  19. 19 MetaVoice 3
  20. 20 VALL-E-X 3
  21. 21 Orpheus-TTS 3
  22. 22 DAC 6kbps 1
  23. 23 SpeechTokenizer 2 layers 1
  24. 24 Encodec 1.5kbps 1

Source: The CodecMOS-Accent paper

#EntryScore ±SettingDateSource
1 CosyVoice 2 4 ±0.1 Official leaderboard
2 OpenAudio s1 mini 4 ±0.1 Official leaderboard
3 Llasa-1B 4 ±0.1 Official leaderboard
4 Chatterbox 4 ±0.1 Official leaderboard
5 NanoCodec 4 ±0.1 Official leaderboard
6 FACodec 4 ±0.1 Official leaderboard
7 NeuTTS Air 4 ±0.1 Official leaderboard
8 NeuCodec 4 ±0.1 Official leaderboard
9 IndexTTS2 4 ±0.1 Official leaderboard
10 TorToiSe 4 ±0.1 Official leaderboard
11 VoiceStar 4 ±0.1 Official leaderboard
12 XTTS 4 ±0.1 Official leaderboard
13 MaskGCT 4 ±0.1 Official leaderboard
14 VevoTTS 4 ±0.1 Official leaderboard
15 FireRedTTS 4 ±0.1 Official leaderboard
16 SNAC 3 ±0.1 Official leaderboard
17 Mimi 4.4kbps 3 ±0.1 Official leaderboard
18 WavTokenizer 40hz 3 ±0.1 Official leaderboard
19 MetaVoice 3 ±0.1 Official leaderboard
20 VALL-E-X 3 ±0.1 Official leaderboard
21 Orpheus-TTS 3 ±0.1 Official leaderboard
22 DAC 6kbps 1 ±0.1 Official leaderboard
23 SpeechTokenizer 2 layers 1 ±0.1 Official leaderboard
24 Encodec 1.5kbps 1 ±0.1 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

All 695 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

7 of 8 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 2 pass 3 notes
  • Pass Independent runs Scores from the benchmark’s paper (24 entries).
  • Pass Models scored 24 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 rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 24 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.

How we check