Best AI Voice & Speech Models
See how the top voice models compare, when each launched, and which labs are ahead.
ELO arena ranking for speech synthesis quality based on blind pairwise human preference comparisons. Data from Artificial Analysis
Top rated
1
Qwen-Audio-3.0-TTS-Plus1,234Best open weights
1
Step TTS 21,105Longest at #1
1
Microsoft Azure1,478 days
Microsoft Azure1,478 days2
Google1,144 days
Google1,144 days3
OpenAI557 days
OpenAI557 days| # | ||
|---|---|---|
| 1 | Alibaba | 1,234±16 |
| 2 |  SpeechifyAI | 1,230±16 |
| 3 | Google | 1,215±13 |
| 4 |  Inworld | 1,212±19 |
| 5 |  Cartesia | 1,208±14 |
| 6 | Alibaba | 1,205±15 |
| 7 | StepFun | 1,199±16 |
| 8 |  Smallest.ai | 1,198±17 |
| 9 | Inworld | 1,198±14 |
| 10 | Inworld | 1,197±14 |
| 11 | VUI Labs | 1,194±16 |
| 12 | Alibaba | 1,193±15 |
| 13 |  MiniMax | 1,178±12 |
| 14 | StepFun | 1,175±18 |
| 15 |  ElevenLabs | 1,175±12 |
| 16 |  async | 1,164±15 |
| 17 | Inworld | 1,161±13 |
| 18 | Inworld | 1,159±16 |
| 19 | MiniMax | 1,150±12 |
| 20 | Smallest.ai | 1,146±15 |
| 21 | StepFun | 1,142±14 |
| 22 | async | 1,140±15 |
| 23 |  Fish Audio | 1,138±15 |
| 24 | MiniMax | 1,138±12 |
| 25 | Inworld | 1,137±13 |
| 26 | MiniMax | 1,128±12 |
| 27 | Microsoft Azure | 1,126±13 |
| 28 |  Speechify | 1,122±14 |
| 29 | Inworld | 1,121±12 |
| 30 | Fish Audio | 1,120±14 |
| 31 | StepFun | 1,113±14 |
| 32 | MiniMax | 1,112±12 |
| 33 |  SpaceXAI | 1,108±21 |
| 34 | StepFun | 1,105±19 |
| 35 | OpenAI | 1,103±12 |
| 36 | ElevenLabs | 1,102±11 |
| 37 | ElevenLabs | 1,102±11 |
| 38 | ElevenLabs | 1,095±11 |
| 39 | Gradium | 1,094±16 |
| 40 | Smallest.ai | 1,090±14 |
| 41 |  Maya Research | 1,089±17 |
| 42 | Gradium | 1,088±15 |
| 43 | ElevenLabs | 1,086±11 |
| 44 | MiniMax | 1,084±12 |
| 45 | Resemble AI | 1,083±14 |
| 46 | OpenAI | 1,083±11 |
| 47 |  Mistral | 1,075±14 |
| 48 | 1,075±13 | |
| 49 | Google | 1,072±12 |
| 50 | Cartesia | 1,070±12 |
| 51 | Fish Audio | 1,068±12 |
| 52 | OpenAI | 1,066±15 |
| 53 |  Xiaomi | 1,064±14 |
| 54 | Speechify | 1,064±14 |
| 55 |  Amazon | 1,064±12 |
| 56 | StepFun | 1,061±18 |
| 57 | MiniMax | 1,061±12 |
| 58 | 1,060±14 | |
| 59 |  Kokoro | 1,057±11 |
| 60 |  Hume AI | 1,056±13 |
| 61 | OpenAI | 1,056±22 |
| 62 |  Hithink | 1,055±15 |
| 63 | Google | 1,054±12 |
| 64 | Maya Research | 1,048±17 |
| 65 | Amazon | 1,047±14 |
| 66 | Maya Research | 1,046±12 |
| 67 | async | 1,046±12 |
| 68 | Fish Audio | 1,045±20 |
| 69 | Google | 1,044±14 |
| 70 | Google | 1,043±14 |
| 71 |  Rime | 1,042±14 |
| 72 | Cartesia | 1,038±12 |
| 73 | 1,037±13 | |
| 74 | Google | 1,036±14 |
| 75 | 1,035±13 | |
| 76 | Speechify | 1,035±12 |
| 77 | Microsoft Azure | 1,029±14 |
| 78 | Microsoft Azure | 1,027±19 |
| 79 | Smallest.ai | 1,025±14 |
| 80 | Hume AI | 1,025±12 |
| 81 | MiniMax | 1,022±12 |
| 82 | Xiaomi | 1,017±15 |
| 83 | Resemble AI | 1,013±12 |
| 84 | Fish Audio | 1,010±12 |
| 85 |  NVIDIA | 1,005±12 |
| 86 | Rime | 1,003±14 |
| 87 |  Zyphra | 1,000±0 |
| 88 |  Murf AI | 972.0±12 |
| 89 |  LMNT | 970.0±13 |
| 90 | Microsoft Azure | 966.0±14 |
| 91 | StepFun | 959.0±10 |
| 92 | Microsoft Azure | 957.0±14 |
| 93 |  OpenVoice | 956.0±14 |
| 94 | NVIDIA | 940.0±15 |
| 95 |  Neuphonic | 934.0±14 |
| 96 | Alibaba | 931.0±15 |
| 97 | Alibaba | 917.0±14 |
| 98 |  Coqui | 916.0±15 |
| 99 | Google | 902.0±12 |
| 100 |  StyleTTS | 890.0±16 |
| 101 | Rime | 885.0±15 |
| 102 | Google | 883.0±12 |
| 103 | Amazon | 881.0±14 |
| 104 | Google | 876.0±12 |
| 105 | Murf AI | 857.0±15 |
| 106 |  Noiz | 853.0±17 |
| 107 |  MetaVoice | 835.0±18 |
| 108 | Amazon | 810.0±16 |
Artificial Analysis Arena data from Artificial Analysis· updated Jul 26, 2026