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.
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
Diarization
Artificial Analysis Diarization Index
Diarization by Dataset
WDER: AA-Diarization Dataset
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 | 13.5% | 79.5 | $1.51 | ||
Fun-Realtime-ASR-preview | 1.7% | 20.7 | $2.82 | ||
Nova 2 Pro | 4.9% | 23.8 | $3.10 | ||
Amazon Transcribe | 4.1% | 39.2 | $6.00 | ||
Universal-3.5 Pro | 3.0% | 107.4 | $3.50 | ||
Universal-3 Pro | 3.1% | 99.9 | $3.50 | ||
Universal, AssemblyAI | 3.8% | 102.6 | $2.50 | ||
MAI-Transcribe-2 | 2.0% | 323.7 | $1.67 | ||
MAI-Transcribe-1.5 | 2.4% | 79.1 | $6.00 | ||
Nova-3 | 5.2% | 389.9 | $4.30 | ||
Nova-3, Telnyx | 4.8% | 340.8 | $5.00 | ||
Nova-2, Telnyx | 5.1% | 330.5 | $5.00 | ||
Scribe v3, ElevenLabs | 1.9% | 74.3 | $3.33 | ||
Scribe v2 | 2.2% | 57.8 | $3.67 | ||
Solaria-1, Gladia | 4.1% | 83.8 | $5.02 | ||
Solaria-3, Gladia | 3.2% | 89.0 | $2.98 | ||
Chirp 3, Google | 4.3% | 29.2 | $16.00 | ||
Chirp | 31.2% | 14.4 | $16.00 | ||
Gemini 3.5 Transcribe | 2.6% | 92.7 | $5.00 | ||
Gemini 3.1 Pro Preview (High) | 2.8% | 8.0 | $18.15 | ||
Gemini 3.1 Pro Preview (Low) | 3.6% | 8.2 | $7.72 | ||
Gemini 3 Flash (High) | 2.9% | 24.3 | $4.17 | ||
Gemini 2.5 Flash Lite | 5.2% | 84.5 | $0.67 | ||
Gemini 2.5 Flash | 5.1% | 71.7 | $2.51 | ||
Gemini 2.5 Pro | 2.9% | 13.1 | $8.69 | ||
Gemini 3.1 Flash-Lite Preview (Minimal) | 3.4% | 68.7 | $1.10 | ||
Gradium Speech-to-Text | 6.8% | 2.8 | $13.00 | ||
Grok Voice Transcribe 2.0 | 2.3% | 186.4 | $1.67 | ||
Inworld STT 1 | 3.9% | 79.2 | $2.50 | ||
Voxtral Small Transcribe 3 | 2.1% | 50.7 | $4.00 | ||
Voxtral Mini Transcribe 3 | 2.2% | 57.8 | $3.50 | ||
Voxtral Mini Transcribe 2 | 3.6% | 86.2 | $3.00 | ||
Voxtral Small | 2.8% | 61.4 | $4.00 | ||
Voxtral Mini | 3.8% | 51.2 | $1.00 | ||
Modulate STT Batch English VFast | 4.2% | 46.5 | $0.42 | ||
Nemotron 3.5 Transcribe | 3.4% | 63.9 | $2.00 | ||
Parakeet TDT 0.6B V3, Togetherai | 4.5% | 665.8 | $1.50 | ||
Canary Qwen 2.5B, NVIDIA | 4.3% | 6.7 | $0.74 | ||
Parakeet TDT 0.6B V2, NVIDIA | 6.4% | 97.4 | $0.00 | ||
Parakeet RNNT 1.1B | 5.4% | 6.3 | $1.91 | ||
GPT Transcribe, OpenAI | 3.3% | 48.4 | $4.50 | ||
GPT-4o Transcribe | 4.0% | 35.0 | $6.00 | ||
GPT-4o Mini Transcribe | 4.5% | 47.9 | $3.00 | ||
Smallest AI Pulse Pro | 3.4% | 294.2 | $4.00 | ||
Resonant-1 | 3.4% | 348.5 | $22.33 | ||
Rev AI | 5.9% | 14.3 | $3.33 | ||
Smallest AI Pulse | 3.3% | 241.8 | $3.00 | ||
Soniox v6 Async | 3.0% | 81.1 | $1.66 | ||
Soniox v5 Async | 3.8% | 47.0 | $1.66 | ||
Speechmatics Melia | 4.9% | 178.1 | $4.00 | ||
Speechmatics Standard | 5.1% | 87.2 | $7.50 | ||
Speechmatics Enhanced | 4.0% | 54.0 | $12.50 | ||
StepAudio 2.5 ASR, StepFun | 4.7% | 85.8 | $0.37 | ||
StepAudio 3 ASR, StepFun | 2.5% | 80.0 | $6.67 | ||
Whisper Large v3 Turbo | 4.6% | 391.7 | $0.67 | ||
Whisper Large v3 Turbo, Telnyx | 5.5% | 36.7 | $15.00 | ||
Wizper Large v3 | 4.7% | 143.7 | $0.50 | ||
Incredibly Fast Whisper | 5.7% | 85.4 | $1.49 | ||
Whisper Large v3 | 10.1% | 2.8 | $4.23 | ||
Whisper Large v3 | 4.1% | 98.7 | $1.15 | ||
Whisper Large v3 | 4.5% | 236.7 | $1.50 | ||
Whisper Large v2 | 4.1% | 26.2 | $6.00 | ||
Zoom Scribe V2.1 Pro | 2.1% | 187.7 | $5.83 | ||
Zoom Scribe V2 Pro | 2.8% | 123.7 | $5.83 |
Frequently Asked Questions
Fun-Realtime-ASR-preview leads with the lowest AA-WER (Artificial Analysis Word Error Rate) of 1.7% across 65 models evaluated.
The top speech to text models by accuracy (AA-WER) are: 1. Fun-Realtime-ASR-preview (1.7%), 2. Scribe v3, ElevenLabs (1.9%), 3. MAI-Transcribe-2 (2.0%), 4. Zoom Scribe V2.1 Pro (2.1%), and 5. Voxtral Small Transcribe 3 (2.1%). Lower AA-WER indicates better transcription accuracy.
Parakeet TDT 0.6B V3, Togetherai is the fastest with a speed factor of 665.8x real-time, followed by Whisper (L, v3, Turbo), Groq (391.7x) and Nova-3 (389.9x). Higher speed factors mean faster transcription.
StepAudio 2.5 ASR is the most affordable at $0.37 per 1,000 minutes, followed by Modulate STT Batch English VFast ($0.42) and Wizper (L, v3), fal.ai ($0.50).
SOV-Transcribe-1, Salesforce is the most accurate open weights model with an AA-WER of 2.3%. There are 16 open weights models out of 65 total evaluated.
The top open weights speech to text models by accuracy are: 1. SOV-Transcribe-1, Salesforce (AA-WER 2.3%), 2. Voxtral Small, Mistral (AA-WER 2.8%), and 3. Nemotron 3.5 Transcribe (AA-WER 2.8%).
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.