Audio Model Rankings

Best TTS & Speech-to-Text Models

Text-to-speech and ASR models ranked by TTS Arena preference, WER, and MOS.

Benchmark data last synced Nov 8, 2026

59 models

		Kind
AA83.3	NVIDIA Parakeet TDT 0.6B v3 NVIDIA	asr	87 Average	0.6B	0.9 GB
AA83.3	NVIDIA Parakeet TDT 0.6B v2 NVIDIA	asr	88 Average	0.6B	0.9 GB
AA81.9	Kokoro v1.0 hexgrad	tts	79 Average	0.082B	0.6 GB
AA81.2	Qwen3-ASR-0.6BRising Alibaba	asr	87 Average	0.6B	0.9 GB
AA80.0	XTTS v2 Coqui	voice-clone	73 Average	0B	0.5 GB
AA77.8	Fish Speech v1.4 Fish Audio	tts	98 Average	0B	0.5 GB
AA76.6	Moonshine Streaming MediumTrending Useful Sensors	asr	87 Average	0.245B	0.7 GB
AA76.4	Fish Speech v1.5 Fish Audio	tts	86 Average	0B	0.5 GB
AdTools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now
AA76.0	NVIDIA Canary 1B v2Trending NVIDIA	asr	86 Average	0.978B	1.1 GB
AA74.8	Granite-Speech-5.0-470M-TurboCTCNew IBM	asr	89 Average	0.47B	0.8 GB
AA74.3	Qwen3-ASR-1.7B Alibaba	asr	88 Average	1.7B	1.5 GB
AA73.6	CosyVoice 2.0Trending Alibaba	tts	91 Average	0.5B	0.8 GB
AA72.7	NVIDIA Canary 180M Flash NVIDIA	asr	86 Average	0.182B	0.6 GB
AA71.8	Cohere Transcribe (03-2026) Cohere	asr	89 Average	2B	1.7 GB
AA71.0	Distil-Whisper Large v3.5Trending Hugging Face	asr	86 Average	0.8B	1.0 GB

Rows per page

Page 1 of 4

Related AI Model Tools

Monthly Report

Which models lead, where API prices moved, and what it costs to run them. Measured from live data.

GPU Guides59