AI Model Leaderboards

AI Speech Synthesis Elo Ranking · China

Speech synthesis models ranked by Artificial Analysis's text-to-speech arena Elo: a relative score from head-to-head battles, comparable only within this board. Synced once a day.

Global Leaderboard China Leaderboard

Intelligence Coding Agentic Image Video Speech

AI Speech Synthesis Elo Ranking· ChinaTOP 21

Updated Sep 30, 2026

ModelDeveloperScore

Qwen: Qwen-Audio-3.0-TTS-PlusQwen-Audio-3.0-TTS-Plus
95% CI ±16
Qwen1,258
StepFun: StepAudio 2.5 TTSStepAudio 2.5 TTS
95% CI ±16
StepFun1,206
MiniMax: Speech 2.8 HDSpeech 2.8 HD
95% CI ±11
MiniMax1,171
4
MiniMax: Speech 2.8 TurboSpeech 2.8 Turbo
95% CI ±11
MiniMax1,150
5
Fish Audio: Fish Audio S2.1 ProFish Audio S2.1 Pro
95% CI ±13
Fish Audio1,139
6
StepFun: Step TTS 2Step TTS 2
95% CI ±13
StepFun1,134
7
MiniMax: Speech 2.6 HDSpeech 2.6 HD
95% CI ±11
MiniMax1,128
8
MiniMax: Speech 2.6 TurboSpeech 2.6 Turbo
95% CI ±11
MiniMax1,119
9
Fish Audio: Fish Audio S2 ProFish Audio S2 Pro
95% CI ±12
Fish Audio1,118
10
MiniMax: Speech-02-HDSpeech-02-HD
95% CI ±11
MiniMax1,112
11
StepFun: Step Audio EditXStep Audio EditX
95% CI ±13
StepFun1,093
12
MiniMax: Speech-02-TurboSpeech-02-Turbo
95% CI ±12
MiniMax1,088
13
Fish Audio: OpenAudio S1OpenAudio S1
95% CI ±11
Fish Audio1,080
14
MiniMax: T2A-01-HDT2A-01-HD
95% CI ±11
MiniMax1,064
15
Xiaomi: MiMo-V2.5-TTSMiMo-V2.5-TTS
95% CI ±13
Xiaomi1,052
16
Fish Audio: OpenAudio S1 MiniOpenAudio S1 Mini
95% CI ±20
Fish Audio1,042
17
MiniMax: T2A-01-TurboT2A-01-Turbo
95% CI ±11
MiniMax1,029
18
Xiaomi: MiMo-V2-TTSMiMo-V2-TTS
95% CI ±14
Xiaomi1,021
19
OpenVoice: OpenVoice v2OpenVoice v2
95% CI ±13
OpenVoice952
20
Qwen: Qwen3 TTS FlashQwen3 TTS Flash
95% CI ±14
Qwen944
21
Qwen: Qwen3 TTSQwen3 TTS
95% CI ±13
Qwen930

Source: Artificial Analysis · arena Elo rating, as of Sep 30, 2026. Elo comes from head-to-head arena battles rather than a fixed exam, so it is a relative score: it only compares within the same board, and it gets steadier as more battles come in.

Arena ratings come from Artificial Analysis's public evaluations and sync once a day; scores for brand-new models move around more until enough battles are in. Names link to the matching tool page on our site.

The China board groups models by where the team is based (Alibaba, ByteDance, Tencent, Baidu, DeepSeek, Moonshot, Zhipu, MiniMax, StepFun and others), not by the model's language or where it is hosted. Everything else counts as overseas, so this board has fewer entries than the global one.

Disclaimer: this board is generated automatically from third-party public data, for reference only, with no warranty or recommendation and no view on any product's safety or legality; names and trademarks belong to their respective owners, and we are not affiliated with or endorsed by these products. To report an error or ask for a correction or removal, use the contact details on our cooperation page to reach us. By using this site you accept our Terms of Service and Privacy Policy, and we accept no liability for any loss arising from use of this board.