Speech-to-Text
Multilingual
Accuracy beyond English — Spanish, German, French and Arabic, each on its own FLEURS-50 split through the same gateway. Ranked by median WER; Arabic by CER. A language shows only the providers we measured on it.
| # | Model | WER · lower ▸ |
|---|---|---|
| 1 | GPT-4o-mini TranscribeOpenAI | 1.2% |
| 2 | Scribe v2ElevenLabs | 1.6% |
| 3 | GPT-4o TranscribeOpenAI | 2.4% |
| 4 | Qwen3-ASRAlibaba | 3.3% |
| 5 | stt-rt-v5Soniox | 3.3% |
| 6 | Grok STTxAI | 3.6% |
| 7 | Scribe v1ElevenLabs | 3.8% |
| 8 | Chirp 3Google | 4.2% |
| 9 | Gradium ASRGradium | 4.3% |
| 10 | Ink-WhisperCartesia | 5.1% |
| 11 | Pulse ProSmallest AI | 6.0% |
| 12 | Nova-3Deepgram | 6.7% |
| 13 | Nova-2Deepgram | 6.7% |
