Speech to Text AI Model & Provider Leaderboard
Compare word error rate, speed, and pricing across Speech to Text models and providers.
For further details, see our methodology page.
Highlights
Artificial Analysis Word Error Rate Index (Non-streaming)
Artificial Analysis Word Error Rate Index (Non-streaming)
AA-WER (Non-streaming) by Dataset
AA-WER (Non-streaming): AA-AgentTalk Dataset
Cleaned Dataset Comparison
VoxPopuli: Cleaned vs Original Subset of Publicly Available Data
API Benchmarks
Artificial Analysis Word Error Rate Index (Non-streaming) vs. Price
Speed Factor
Price of Transcription
Summary of Key Metrics & Further Information
Provider | Further Details | ||||
|---|---|---|---|---|---|
Qwen3.5 Omni Flash |  Alibaba Cloud | 13.5% | 73.4 | 0.00 | |
Qwen3.5 Omni Plus | Alibaba Cloud | 3.5% | 97.4 | 0.00 | |
Nova 2 Pro |  Amazon Bedrock | 4.9% | 22.5 | 3.10 | |
Amazon Transcribe | Amazon Bedrock | 4.1% | 19.0 | 24.00 | |
Universal-3 Pro |  AssemblyAI | 3.1% | 83.6 | 3.50 | |
Universal, AssemblyAI | AssemblyAI | 3.8% | 114.4 | 2.50 | |
MAI-Transcribe-1.5 |  Microsoft Azure | 2.4% | 202.0 | 6.00 | |
MAI-Transcribe-1 | Microsoft Azure | 2.6% | 56.2 | 6.00 | |
transcribe-03-2026 |  Cohere | 4.6% | 64.2 | 0.00 | |
Nova-3 |  Deepgram | 5.2% | 504.4 | 4.30 | |
Nova-2 | Deepgram | 5.3% | 479.2 | 4.30 | |
Base | Deepgram | 10.7% | 353.3 | 12.50 | |
Scribe v2 |  ElevenLabs | 2.2% | 34.0 | 3.67 | |
Scribe v1 | ElevenLabs | 3.0% | 39.3 | 6.67 | |
Solaria-1, Gladia |  Gladia | 4.1% | 60.6 | 4.07 | |
Gemini 3.1 Pro Preview (High) |  Google | 2.8% | 6.8 | 18.15 | |
Gemini 3.1 Pro Preview (Low) | Google | 3.6% | 6.5 | 7.72 | |
Gemini 3 Flash (High) | Google | 2.9% | 16.9 | 13.70 | |
Gemini 2.5 Flash Lite | Google | 5.2% | 69.3 | 6.56 | |
Gemini 2.5 Flash | Google | 5.1% | 69.4 | 6.66 | |
Gemini 2.5 Pro | Google | 2.9% | 12.3 | 11.39 | |
Gemini 3.1 Flash-Lite Preview (Minimal) | Google | 3.4% | 75.4 | 5.83 | |
Gradium Speech-to-Text |  Gradium | 8.4% | 2.3 | 13.00 | |
Grok Speech to Text, xAI |  xAI | 4.0% | 100.5 | 1.67 | |
LLM Speech, Azure | Microsoft Azure | 3.7% | 57.5 | 6.00 | |
Voxtral Mini Transcribe 2 |  Mistral | 3.6% | 73.4 | 3.00 | |
Voxtral Mini Transcribe | Mistral | 3.5% | 72.7 | 2.00 | |
Voxtral Small | Mistral | 2.8% | 66.1 | 4.00 | |
Voxtral Mini |  DeepInfra | 3.8% | 74.7 | 1.00 | |
Modulate STT Batch English VFast |  Modulate | 13.0% | 188.9 | 0.42 | |
Parakeet TDT 0.6B V3, Togetherai |  Together.ai | 4.5% | 991.4 | 1.50 | |
Canary Qwen 2.5B, NVIDIA |  Replicate | 4.3% | 5.8 | 0.74 | |
Parakeet TDT 0.6B V2, NVIDIA |  NVIDIA | 6.4% | 101.2 | 0.00 | |
Parakeet RNNT 1.1B | Replicate | 5.4% | 6.3 | 1.91 | |
GPT-4o Transcribe |  OpenAI | 4.0% | 33.6 | 6.00 | |
GPT-4o Mini Transcribe | OpenAI | 4.5% | 48.3 | 3.00 | |
Rev AI |  Rev AI | 5.9% | 12.9 | 3.33 | |
Smallest AI Pulse |  Smallest.ai | 4.4% | 135.3 | 5.00 | |
Speechmatics Standard |  Speechmatics | 5.1% | 67.5 | 4.00 | |
Speechmatics Enhanced | Speechmatics | 4.0% | 52.3 | 6.70 | |
Whisper Large v3 Turbo |  Groq | 4.6% | 119.7 | 0.67 | |
Whisper Large v3 Turbo |  Fireworks | 4.7% | 161.9 | 1.00 | |
Wizper Large v3 |  fal.ai | 4.7% | 201.5 | 0.50 | |
Incredibly Fast Whisper | Replicate | 5.7% | 54.3 | 1.49 | |
Whisper Large v3 | Replicate | 10.1% | 2.8 | 4.23 | |
Whisper Large v3 | fal.ai | 4.1% | 98.8 | 1.15 | |
Whisper Large v3 | Fireworks | 4.6% | 324.6 | 1.00 | |
Whisper Large v3 | Together.ai | 4.5% | 428.0 | 1.50 | |
Whisper Large v2 | OpenAI | 4.1% | 26.8 | 6.00 |
Frequently Asked Questions
Fun-Realtime-ASR-preview leads with the lowest AA-WER (Artificial Analysis Word Error Rate) of 1.7% across 54 models evaluated.
The top speech to text models by accuracy (AA-WER) are: 1. Fun-Realtime-ASR-preview (1.7%), 2. Scribe v2, ElevenLabs (2.2%), 3. MAI-Transcribe-1.5 (2.4%), 4. MAI-Transcribe-1 (2.6%), 5. Gemini 3 Pro (High), Google (2.7%). Lower AA-WER indicates better transcription accuracy.
Parakeet TDT 0.6B V3, Togetherai is the fastest with a speed factor of 991.4x real-time, followed by Nova-3 (504.4x) and Nova-2 (479.2x). Higher speed factors mean faster transcription.
Modulate STT Batch English VFast is the most affordable at $0.417 per 1,000 minutes, followed by Wizper (L, v3), fal.ai ($0.50) and Whisper (L, v3, Turbo), Groq ($0.667).
Voxtral Small, Mistral is the most accurate open weights model with an AA-WER of 2.8%. There are 13 open weights models out of 54 total evaluated.
The top open weights speech to text models by accuracy are: 1. Voxtral Small, Mistral (AA-WER 2.8%), 2. Voxtral Mini Transcribe, Mistral (AA-WER 3.5%), 3. Voxtral Mini Transcribe 2, Mistral (AA-WER 3.6%).
The best model depends on your priorities. Use the scatter plots to visualize trade-offs between accuracy (AA-WER), speed, and price. For applications requiring high accuracy, prioritize models with lower AA-WER scores. For real-time applications, focus on speed factor. For cost-sensitive workloads, compare the price charts.
Speech to Text models & providers compared: MAI-Transcribe-1.5, MAI-Transcribe-1, Grok Speech to Text, xAI, Qwen3.5 Omni Flash, Qwen3.5 Omni Plus, Cohere Transcribe 03-2026, Gemini 3.1 Pro Preview (High), Gemini 3.1 Pro Preview (Low), Smallest AI Pulse, Voxtral Mini Transcribe 2, Universal-3 Pro, Scribe v2, Gemini 3 Flash (High), Nova 2 Pro, Gradium Speech-to-Text, LLM Speech, Azure, Parakeet TDT 0.6B V3, Togetherai, Gemini 2.5 Flash Lite, Canary Qwen 2.5B, Replicate, Voxtral Mini Transcribe, Voxtral Small, Voxtral Mini, Deepinfra, Gemini 2.5 Flash, Gemini 2.5 Pro, Parakeet TDT 0.6B V2, Solaria-1, GPT-4o Transcribe, GPT-4o Mini Transcribe, Scribe v1, Nova-3, Universal, Whisper (L, v3, Turbo), Groq, Whisper (L, v3, Turbo), Fireworks, Parakeet RNNT 1.1B, Replicate, Amazon Transcribe, Wizper (L, v3), fal.ai, Incredibly Fast, Replicate, Whisper (L, v3), Replicate, Whisper (L, v3), fal.ai, Whisper (L, v3), Fireworks, Whisper Large v3, together.ai, Nova-2, Standard, Enhanced, Large v2, Base, Rev AI, Gemini 3.1 Flash-Lite Preview (Minimal), Modulate STT Batch English VFast.