Voice & Audio Intelligence
Speech-to-Text & Text-to-Speech Arena
Standardized evaluation across the AA-WER v2 composite index (AgentTalk, VoxPopuli, Earnings22), voice naturalness Elo, real-time factor (RTF), and streaming latency.
Speech-to-Text (ASR)
| # | Model & Provider | AA-WER Error Rate (Lower is Better) | Speed Factor (RTF) | Time to First Partial | Cost / 1,000 Minutes | Streaming Support |
|---|---|---|---|---|---|---|
| 1 | Nova-3 (Speech-to-Text) Deepgram | 4.8% | 110x speed | 0.18s | $4.30 | Streaming & Batch |
| 2 | Scribe v2 ElevenLabs | 4.9% | 92x speed | 0.2s | $7.50 | Streaming & Batch |
| 3 | Universal-3.5-Pro AssemblyAI | 5.1% | 95x speed | 0.22s | $6.50 | Streaming & Batch |
| 4 | Flux Deepgram | 5.1% | 130x speed | 0.14s | $5.90 | Streaming |
| 5 | Gemini 3.5 Transcribe Pro Google Cloud | 5.2% | 75x speed | 0.25s | $7.20 | Streaming & Batch |
| 6 | Soniox STT Soniox | 5.3% | 105x speed | 0.19s | $3.90 | Streaming & Batch |
| 7 | Whisper large-v3 turbo OpenAI / Open Source | 5.4% | 85x speed | 0.35s | $6.00 | Batch Async |
| 8 | Speechmatics Ursa Speechmatics | 5.5% | 88x speed | 0.24s | $5.40 | Streaming & Batch |
| 9 | Scribe v1 (Conversational) ElevenLabs | 5.6% | 60x speed | 0.28s | $8.00 | Streaming & Batch |
| 10 | Grok STT xAI | 5.7% | 98x speed | 0.21s | $5.00 | Streaming & Batch |
| 11 | GPT-4o Transcribe OpenAI | 5.8% | 55x speed | 0.32s | $6.00 | Batch Async |
| 12 | Solaria-1 Gladia | 5.9% | 78x speed | 0.27s | $6.10 | Streaming & Batch |
| 13 | Qwen3-ASR Flash Alibaba Cloud | 6.1% | 115x speed | 0.22s | $3.40 | Streaming & Batch |
| 14 | Nemotron ASR NVIDIA | 6.2% | 90x speed | 0.3s | $4.80 | Streaming & Batch |
| 15 | Ink-Whisper Cartesia | 6.3% | 120x speed | 0.17s | $2.90 | Streaming |
| 16 | Cohere Transcribe Cohere | 6.4% | 70x speed | 0.29s | $5.60 | Streaming & Batch |
| 17 | Parakeet TDT v3 NVIDIA | 6.5% | 140x speed | 0.15s | $1.80 | Streaming & Batch |
| 18 | Mistral Voxtral STT Mistral AI | 6.7% | 80x speed | 0.31s | $4.20 | Batch Async |
| 19 | Distil-Whisper large-v3 Hugging Face | 6.9% | 150x speed | 0.2s | $1.20 | Batch Async |
| 20 | Whisper v1 (API) OpenAI | 7.2% | 45x speed | 0.4s | $6.00 | Batch Async |
Text-to-Speech (TTS)
| # | Model & Provider | Naturalness Arena Elo | Time to First Audio | Cost / 1M Characters | Voices & Languages | License |
|---|---|---|---|---|---|---|
| 1 | Eleven v3 ElevenLabs | 1288 | 250ms | $180.00 | 5,000+ / 70+ langs | Proprietary |
| 2 | Gemini 3.1 TTS Google | 1264 | 210ms | $30.00 | 30+ / 24 langs | Proprietary |
| 3 | Sonic 3.6 Cartesia | 1249 | 90ms | $65.00 | 100+ / 42 langs | Proprietary |
| 4 | GPT Realtime 2 (Speech) OpenAI | 1238 | 160ms | $48.00 | 12 / 50+ langs | Proprietary |
| 5 | Speech-2.8-HD MiniMax | 1227 | 190ms | $60.00 | 300+ / 40 langs | Proprietary |
| 6 | PlayAI 3.0 Dialog PlayAI | 1215 | 140ms | $50.00 | 200+ / 30 langs | Proprietary |
| 7 | Eleven Turbo v2.5 ElevenLabs | 1209 | 120ms | $90.00 | 5,000+ / 32 langs | Proprietary |
| 8 | Sonic 3.5 Cartesia | 1202 | 95ms | $55.00 | 100+ / 40 langs | Proprietary |
| 9 | Maya 2 Maya Research | 1196 | 180ms | $45.00 | 50+ / 20 langs | Proprietary |
| 10 | Speech-2.6-HD MiniMax | 1190 | 200ms | $55.00 | 300+ / 35 langs | Proprietary |
| 11 | Qwen3-TTS Flash Alibaba Cloud | 1183 | 130ms | $15.00 | 17 / 10 langs | Proprietary |
| 12 | Step-Audio TTS StepFun | 1177 | 170ms | $25.00 | 60+ / 8 langs | Proprietary |
| 13 | TTS-1 HD OpenAI | 1170 | 240ms | $30.00 | 6 / 50+ langs | Proprietary |
| 14 | VibeVoice Microsoft | 1165 | 220ms | $12.00 | Multi-speaker / EN-ZH | Open Weights |
| 15 | Dia 1.6B Nari Labs | 1158 | 300ms | $8.00 | Voice clone / EN | Open Weights |
| 16 | Kokoro 82M Hexgrad | 1155 | 60ms | $3.00 | 54 / 9 langs | Apache-2.0 |
| 17 | OpenAudio S1 Fish Audio | 1150 | 150ms | $20.00 | Voice clone / 13 langs | Open Weights |
| 18 | Voxtral TTS Mistral AI | 1146 | 175ms | $22.00 | 8 / 12 langs | Open Weights |
| 19 | XTTS v2 Coqui | 1140 | 260ms | $5.00 | Voice clone / 17 langs | Coqui CPML |
| 20 | Zonos v0.1 Zyphra | 1135 | 280ms | $10.00 | Voice clone / 5 langs | Apache-2.0 |
| 21 | Chatterbox Resemble AI | 1129 | 310ms | $6.00 | Voice clone / EN | MIT |
| 22 | Orpheus 3B Canopy Labs | 1121 | 200ms | $9.00 | 8 / EN | Apache-2.0 |
| 23 | Amazon Polly (Generative) AWS | 1110 | 140ms | $30.00 | 100+ / 40 langs | Proprietary |