Audio AI Benchmarks

Blind-vote Elo for text-to-speech and word error rate for speech-to-text, from the Artificial Analysis speech leaderboards (refreshed 2026-11-15).

Showing 107 of 107 models

#ModelLabEloVotes$ / 1M charsReleased
1Eleven v4ElevenLabs1340 ±152,742$80.00Sept 2026
2Eleven v4 TurboElevenLabs1339 ±162,446$40.00Sept 2026
3Qwen-Audio-3.1-TTS-PlusAlibaba Cloud / Qwen1300 ±142,490$19.31Sept 2026
4Sonic 3.6Cartesia1294 ±133,366$49.00Aug 2026
5Simba 3.5Speechify1294 ±171,732$6.58Nov 2026
6Gemini 3.8 Flash TTSGoogle DeepMind1284 ±133,633$16.49Sept 2026
7Fish Audio Drama 3Fish Audio1275 ±171,742$15.00Oct 2026
8Lightning V3.1 Pro (Oct 2026)Smallest.ai1273 ±152,064$19.50Oct 2026
9Realtime TTS-2Inworld1268 ±132,761$20.83Aug 2026
10Qwen-Audio-3.0-TTS-PlusAlibaba Cloud / Qwen1264 ±133,056$19.31Sept 2026
11Gemini 3.8 Flash-Lite TTSGoogle DeepMind1260 ±123,640$11.03Sept 2026
12Simba 3.2Speechify1250 ±113,966$6.58Jul 2026
13Fish Audio Drama 3 PreviewFish Audio1239 ±132,795$15.00Oct 2026
14Breeze TTS 2BreezeBlue1229 ±132,937$34.00Aug 2026
15StepAudio 2.5 TTS (Aug 2026)StepFun1215 ±122,814$85.00Aug 2026
16AssemblyAI-TTS-1AssemblyAI1215 ±151,867$15.00Oct 2026
17Realtime TTS-2 FlashInworld1215 ±123,235$10.42Aug 2026
18v3 ConversationalElevenLabs1214 ±123,358$50.00Aug 2026
19Gemini 3.1 Flash TTSGoogle DeepMind1213 ±104,812$18.31Apr 2026
20Voxtral 3 TTSMistral AI1213 ±132,258$13.00Oct 2026
21Luna TTSVUI Labs1199 ±114,052$15.00Jun 2026
22Sonic 3.5Cartesia1195 ±114,319$49.00May 2026
23Eleven v3ElevenLabs1189 ±105,557$100.00Feb 2026
24Fish Audio S2.1 Pro (Oct 2026)Fish Audio1185 ±132,635$15.00Oct 2026
25Ming-3.0-ttsAnt Group1182 ±132,640—Oct 2026
26Soniox TTS Real-Time v2Soniox1181 ±105,924$14.23Aug 2026
27Gradium TTS (Sep 2026)Gradium1175 ±122,740$47.78Sept 2026
28Speech 2.8 HDMiniMax1175 ±105,648$100.00Feb 2026
29Falcon 2.1Murf AI1175 ±25236$10.00Oct 2026
30Gradium TTS (Oct 2026)Gradium1168 ±132,147$47.78Sept 2026
31SpaceXAI TTSxAI1161 ±122,298$15.00Mar 2026
32Falcon 2Murf AI1158 ±122,913$10.00Jul 2026
33Gradium TTS (Aug 2026)Gradium1153 ±122,965$47.22Aug 2026
34Speech 2.8 TurboMiniMax1151 ±105,391$60.00Feb 2026
35Azure HD 2.5Microsoft Research & Azure1145 ±113,350$22.00Mar 2026
36Async Flash v1.5async1138 ±113,082$10.10May 2026
37Step TTS 2 (Mar 2026)StepFun1136 ±113,137$40.00Feb 2026
38Async Pro v1.0async1128 ±112,931$20.20May 2026
39Speech 2.6 HDMiniMax1124 ±104,616$100.00Oct 2025
40Speech 2.6 TurboMiniMax1123 ±104,850$60.00Oct 2025
41Simba 3.0Speechify1114 ±113,340$6.58Feb 2026
42Phonon 31M (English)Gradium1111 ±131,943$0.00Oct 2026
43Coda v2Rime1111 ±131,957$50.00Oct 2026
44Chatterbox HDResemble AI1111 ±112,962$40.00May 2025
45TTS-1 HDOpenAI1108 ±114,052$30.00Nov 2023
46Speech-02-HDMiniMax1104 ±114,908$100.00Mar 2025
47Fish Audio S2 ProFish Audio1104 ±112,999$15.00Mar 2026
48Step Audio EditX (Mar 2026)StepFun1100 ±122,607—Feb 2026
49Voxtral TTSMistral AI1094 ±112,744$16.00Mar 2026
50Turbo v2.5ElevenLabs1094 ±107,281$50.00Jul 2024
51Multilingual v2ElevenLabs1093 ±107,702$100.00Aug 2023
52Speech 2.6Hithink1089 ±122,825—Jul 2026
53Speech-02-TurboMiniMax1089 ±114,249$60.00Mar 2025
54OpenAudio S1Fish Audio1089 ±114,863$15.00Jun 2025
55TTS-1OpenAI1084 ±106,521$15.00Nov 2023
56Gemini 2.5 Flash Lite TTSGoogle DeepMind1083 ±113,520$9.15Dec 2025
57Flash v2.5ElevenLabs1074 ±106,036$50.00Dec 2024
58StudioGoogle DeepMind1073 ±114,801$160.00Feb 2023
59T2A-01-HDMiniMax1073 ±115,258$50.00Jan 2025
60GPT-Realtime-2OpenAI1072 ±122,554$191.56May 2026
61Gemini 2.5 Flash TTS (Dec 2025)Google DeepMind1070 ±123,131$9.15Dec 2025
62JourneyGoogle DeepMind1068 ±122,528$160.00Dec 2023
63Polly GenerativeAmazon AWS Bedrock1063 ±113,994$30.00May 2024
64Maya 2 FlashMaya Research1062 ±122,449—Jul 2026
65Chirp 3: HDGoogle DeepMind1061 ±113,636$30.00Mar 2025
66Kokoro 82M v1.0Kokoro1061 ±115,628$0.65Jan 2025
67Magpie-Multilingual 357M (Feb 2026)NVIDIA AI & NIM1060 ±122,513—Feb 2026
68Sonic 3Cartesia1059 ±113,754$49.00Oct 2025
69Octave 2Hume AI1057 ±123,382$87.50Oct 2025
70CodaRime1052 ±122,711$50.00May 2026
71Flux TTSDeepgram1052 ±122,475$45.00Aug 2026
72MiMo-V2.5-TTSXiaomi1050 ±122,322——
73Async Flash v1.0async1047 ±114,246$10.10Jul 2025
74Maya 2 GlobalMaya Research1047 ±132,237—Jul 2026
75Maya1Maya Research1046 ±123,502—Jun 2025
76OpenAudio S1 MiniFish Audio1042 ±201,678$15.00Jun 2025
77Octave TTSHume AI1039 ±114,856$87.50Feb 2025
78Higgs Audio V3 TTSBoson AI1038 ±122,394$15.00Jun 2026
79Gemini 2.5 Pro (Dec 2025)Google DeepMind1037 ±122,622$18.31Dec 2025
80Bland Speech v3Bland AI1035 ±122,818$40.00Aug 2026
81Azure NeuralMicrosoft Research & Azure1032 ±201,156$15.00Sept 2018
82Polly Long-FormAmazon AWS Bedrock1026 ±132,106$100.00Nov 2023
83MAI-Voice-1Microsoft Research & Azure1025 ±122,494—Apr 2026
84T2A-01-TurboMiniMax1025 ±115,136$30.00Jan 2025
85Lightning v3.1Smallest.ai1022 ±122,551$25.00Mar 2026
86ChatterboxResemble AI1021 ±114,843$25.00May 2025
87Raon SpeechLMKrafton1002 ±132,432—Apr 2026
88Zonos-v0.1Zyphra1000 anchor4,985$20.00Feb 2025
89Murf Speech Gen 2Murf AI984 ±125,135$100.00Jul 2024
90OpenVoice v2OpenVoice955 ±133,156$8.33Apr 2024
91VibeVoice 1.5BMicrosoft Research & Azure954 ±142,226—Aug 2025
92Neuphonic TTSNeuphonic941 ±142,551$20.79Oct 2024
93Qwen3 TTS FlashAlibaba Cloud / Qwen940 ±142,272$11.03Sept 2025
94Qwen3 TTSAlibaba Cloud / Qwen928 ±133,202$13.79Jan 2026
95Kugel 3KugelAudio925 ±134,583—Jun 2026
96WaveNetGoogle DeepMind916 ±126,466$4.00Sept 2016
97XTTS v2Coqui914 ±152,753$40.44Nov 2023
98Mist V2Rime903 ±153,118$31.13Mar 2025
99Polly NeuralAmazon AWS Bedrock892 ±143,022$16.00Jul 2019
100Neural2Google DeepMind890 ±126,736$16.00Jun 2022
101StyleTTS 2StyleTTS888 ±152,753$2.82Jun 2023
102StandardGoogle DeepMind884 ±136,822$4.00Aug 2018
103Noiz TTSNoiz866 ±152,529—Nov 2025
104MetaVoice v1MetaVoice846 ±181,710—Feb 2024
105Polly StandardAmazon AWS Bedrock829 ±163,645$4.00Nov 2016
—Bulbul V4 Flashvendor-reportedSarvam AI———2026-10-13
—Bulbul V3vendor-reportedSarvam AI———Feb 2026
107text-to-speech models
67speech-to-text models
56audio labs and providers
Eleven v4top text-to-speech, Elo 1340ElevenLabs
Fun-Realtime-ASR-previewlowest word error rate, 1.7%Alibaba Cloud

How the audio benchmarks work

Text-to-Speech

Speech Arena (Text-to-Speech)

Elo rating (higher is better)

Models are ranked with an Elo system derived from blind user votes: listeners hear two speech samples generated from the same text and pick the one that sounds more natural. The provider-voice leaderboard uses each provider's own native voices. Each result is reported with a 95% interval and the number of votes (samples).

Source: Artificial Analysis
Speech-to-Text

AA-WER v2 (Speech-to-Text)

Word error rate, % of words transcribed incorrectly (lower is better)

Artificial Analysis Word Error Rate index for non-streaming transcription. AA-WER v2 combines three datasets: AA-AgentTalk (50%), VoxPopuli-Cleaned-AA (25%) and Earnings22-Cleaned-AA (25%). Measured independently on dedicated hardware.

Source: Artificial Analysis
Speech-to-Text

Speed Factor (Speech-to-Text)

Input audio seconds transcribed per second (higher is better)

Median speed factor: how many seconds of input audio are transcribed per second of wall-clock time. Measured independently by Artificial Analysis.

Source: Artificial Analysis

Audio labs & providers