Artificial Analysis Speech Arena what it measures, how it is scored and who leads it

Name: Artificial Analysis Speech Arena
Creator: Artificial Analysis

Speech · Artificial Analysis · Arena rating

Leaderboard

Scores from Artificial Analysis Speech Arena, published Nov 11, 2026

Rank	Model	Arena rating
1	Eleven V4ElevenLabs · as “Eleven v4”	1,338
2	Eleven V4 TurboElevenLabs · as “Eleven v4 Turbo”	1,337
3	Qwen Audio 3.1 TTS PlusAlibaba · as “Qwen-Audio-3.1-TTS-Plus”	1,297
4	Sonic 3.6Cartesia · as “Sonic 3.6”	1,296
5	Simba 3.5Speechifyai · as “Simba 3.5”	1,293
6	Gemini 3.8 Flash TTSGoogle · as “Gemini 3.8 Flash TTS”	1,283
7	Fish Audio Drama 3Fish Audio · as “Fish Audio Drama 3”	1,274
8	Lightning V3.1 ProSmallest Ai · as “Lightning V3.1 Pro (Oct 2026)”	1,272
9	Realtime TTS 2Inworld · as “Realtime TTS-2”	1,266
10	Qwen Audio 3.0 TTS PlusAlibaba · as “Qwen-Audio-3.0-TTS-Plus”	1,266
11	Gemini 3.8 Flash Lite TTSGoogle · as “Gemini 3.8 Flash-Lite TTS”	1,259
12	Simba 3.2Speechifyai · as “Simba 3.2”	1,251
13	Breeze TTS 2Breezeblue · as “Breeze TTS 2”	1,229
14	Assemblyai TTS 1AssemblyAI · as “AssemblyAI-TTS-1”	1,222
15	V3 ConversationalElevenLabs · as “v3 Conversational”	1,214
16	Realtime TTS 2 FlashInworld · as “Realtime TTS-2 Flash”	1,213
17	Voxtral 3 TTSMistral AI · as “Voxtral 3 TTS”	1,213
18	Gemini 3.1 Flash TTSGoogle · as “Gemini 3.1 Flash TTS”	1,213
19	Stepaudio 2.5 TTSStepFun · as “StepAudio 2.5 TTS (Aug 2026)”	1,211
20	Luna TTSVui Labs · as “Luna TTS”	1,201
21	Sonic 3.5Cartesia · as “Sonic 3.5”	1,194
22	Eleven V3ElevenLabs · as “Eleven v3”	1,189
23	Ming 3.0 TTSAnt Group · as “Ming-3.0-tts”	1,185
24	Fish Audio S2.1 ProFish Audio · as “Fish Audio S2.1 Pro (Oct 2026)”	1,184
25	Soniox TTS Real Time V2Soniox · as “Soniox TTS Real-Time v2”	1,179

About this benchmark

Task: Read text aloud in a voice people prefer.
Dataset: Blind votes between two voices reading the same text.
Method: Pairwise preference votes turned into an Elo-style rating.
Metric: Arena rating
Organization: Artificial Analysis
Official leaderboard: artificialanalysis.ai/text-to-speech/leaderboard/provider-voice

Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.