Benchmark Almanac

DesignArena Audio Realism Bench

Humanity 1,484 tops its leaderboard Grok TTS 1,179 leads the models we track
Not in model cards

Voice models produce conversational speech, and users vote on which sounds more real. DesignArena built it because 1 to 5 opinion scores for TTS had saturated.

DesignArena introduced it in the post Audio Realism Benchmark: Measuring Realism in Audio Models. It reports artificial speech and human speech views. One of 40 DesignArena boards. DesignArena ranks entries with Bradley-Terry ratings from user votes.

Kind
Audio understanding and voice
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Tasks
All public
Last model added
October 2026
Scores refreshed
November 12th, 2026
33
Entries on its own leaderboard, 4 models we track

Humanity tops the leaderboard with 1,484

Best score per entry from the benchmark's own leaderboard, read on November 12th, 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 Grok TTS, at 1,179.

33 entries. Scroll sideways for the rest. Bars start at 600.

  1. 1,484 Humanity
  2. 1,442 Freya TTS Adam V1
  3. 1,380 Kalpa TTS Beta V0.1
  4. 1,379 Bland Speech v3
  5. 1,341 Cartesia Sonic 3.6
  6. 1,310 Deepgram Flux TTS
  7. 1,238 Eleven v3 Conversational
  8. 1,232 Murf Falcon 2
  9. 1,218 Cartesia Sonic 3.5
  10. 1,208 MAI-Voice-2
  11. 1,185 HeyGen Voice
  12. 1,179 Grok TTS
  13. 1,137 MAI-Voice-2-Flash
  14. 1,135 MiniMax Speech 2.8 HD
  15. 1,121 MiniMax Speech-02 HD
  16. 1,110 Google Gemini 2.5 Pro TTS
  17. 1,103 OpenAI GPT Realtime 2
  18. 1,080 Google Gemini 3.8 Flash TTS
  19. 1,064 OpenAI GPT Realtime 2.1
  20. 1,031 Eleven v4
  21. 1,029 GPT-4o Mini TTS
  22. 1,025 Google Gemini 3.1 Flash TTS
  23. 1,019 Voxtral Mini TTS 3
  24. 998 Eleven v4 Turbo
  25. 970 Eleven v3
  26. 970 Google Gemini 2.5 Flash TTS
  27. 958 Fish Audio Drama 3
  28. 955 Lightning v3.1 Pro
  29. 953 Fish Audio S2.1 Pro
  30. 875 Magpie TTS Multilingual
  31. 860 Inworld Realtime TTS-2
  32. 854 Qwen-Audio-3.0 TTS Flash
  33. 789 Inworld TTS-1.5 Max

Bars start at 600.

  1. 1 Humanity 1,484
  2. 2 Freya TTS Adam V1 1,442
  3. 3 Kalpa TTS Beta V0.1 1,380
  4. 4 Bland Speech v3 1,379
  5. 5 Cartesia Sonic 3.6 1,341
  6. 6 Deepgram Flux TTS 1,310
  7. 7 Eleven v3 Conversational 1,238
  8. 8 Murf Falcon 2 1,232
  9. 9 Cartesia Sonic 3.5 1,218
  10. 10 MAI-Voice-2 1,208
  11. 11 HeyGen Voice 1,185
  12. 12 Grok TTS 1,179
  13. 13 MAI-Voice-2-Flash 1,137
  14. 14 MiniMax Speech 2.8 HD 1,135
  15. 15 MiniMax Speech-02 HD 1,121
  16. 16 Google Gemini 2.5 Pro TTS 1,110
  17. 17 OpenAI GPT Realtime 2 1,103
  18. 18 Google Gemini 3.8 Flash TTS 1,080
  19. 19 OpenAI GPT Realtime 2.1 1,064
  20. 20 Eleven v4 1,031
  21. 21 GPT-4o Mini TTS 1,029
  22. 22 Google Gemini 3.1 Flash TTS 1,025
  23. 23 Voxtral Mini TTS 3 1,019
  24. 24 Eleven v4 Turbo 998
  25. 25 Eleven v3 970
  26. 26 Google Gemini 2.5 Flash TTS 970
  27. 27 Fish Audio Drama 3 958
  28. 28 Lightning v3.1 Pro 955
  29. 29 Fish Audio S2.1 Pro 953
  30. 30 Magpie TTS Multilingual 875
  31. 31 Inworld Realtime TTS-2 860
  32. 32 Qwen-Audio-3.0 TTS Flash 854
  33. 33 Inworld TTS-1.5 Max 789

Source: Official DesignArena Audio Realism Bench leaderboard

#EntryScoreSettingDateSource
1 Humanity 1,484 Official leaderboard
2 Freya TTS Adam V1 1,442 Official leaderboard
3 Kalpa TTS Beta V0.1 1,380 Official leaderboard
4 Bland Speech v3 1,379 Official leaderboard
5 Cartesia Sonic 3.6 1,341 Official leaderboard
6 Deepgram Flux TTS 1,310 Official leaderboard
7 Eleven v3 Conversational 1,238 Official leaderboard
8 Murf Falcon 2 1,232 Official leaderboard
9 Cartesia Sonic 3.5 1,218 Official leaderboard
10 MAI-Voice-2 1,208 Official leaderboard
11 HeyGen Voice 1,185 Official leaderboard
12 Grok TTS 1,179 Apr 2026 Official leaderboard
13 MAI-Voice-2-Flash 1,137 Official leaderboard
14 MiniMax Speech 2.8 HD 1,135 Jan 2026 Official leaderboard
15 MiniMax Speech-02 HD 1,121 Official leaderboard
16 Google Gemini 2.5 Pro TTS 1,110 Official leaderboard
17 OpenAI GPT Realtime 2 1,103 Official leaderboard
18 Google Gemini 3.8 Flash TTS 1,080 Official leaderboard
19 OpenAI GPT Realtime 2.1 1,064 Official leaderboard
20 Eleven v4 1,031 Official leaderboard
21 GPT-4o Mini TTS 1,029 Mar 2025 Official leaderboard
22 Google Gemini 3.1 Flash TTS 1,025 Official leaderboard
23 Voxtral Mini TTS 3 1,019 Oct 2026 Official leaderboard
24 Eleven v4 Turbo 998 Official leaderboard
25 Eleven v3 970 Official leaderboard
26 Google Gemini 2.5 Flash TTS 970 Official leaderboard
27 Fish Audio Drama 3 958 Official leaderboard
28 Lightning v3.1 Pro 955 Official leaderboard
29 Fish Audio S2.1 Pro 953 Official leaderboard
30 Magpie TTS Multilingual 875 Official leaderboard
31 Inworld Realtime TTS-2 860 Official leaderboard
32 Qwen-Audio-3.0 TTS Flash 854 Official leaderboard
33 Inworld TTS-1.5 Max 789 Official leaderboard

3 models have held number one on the official leaderboard since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI MiniMax xAI
Own leaderboard 3 leaders
Grok TTS 1,179

Source: Official DesignArena Audio Realism Bench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
1,000 1,050 1,100 1,150 1,200 2025 2026 2027 1,000 1,050 1,100 1,150 1,200 2025 2026 2027

Source: Official DesignArena Audio Realism Bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Audio and video

All 233 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%