Speech model leaderboard
Compare models within each evaluation task using original scores and configurations.
Ranked by the upstream word error rate index; lower is better. Provider configurations are shown separately.
| 1 | Grok Voice Transcribe 2.0 xAI | 2.29 WER % | Grok Voice Transcribe 2.0 · SpaceXAI | — | 2026-11-05 |
| 2 | Gemini 3.1 Pro Preview Google | 2.82 WER % | Gemini 3.1 Pro Preview (High) · Google | — | 2026-11-05 |
| 3 | Gemini 2.5 Pro Google | 2.86 WER % | Gemini 2.5 Pro · Google | — | 2026-11-05 |
| 4 | Gemini 3.1 Pro Preview Google | 3.58 WER % | Gemini 3.1 Pro Preview (Low) · Google | — | 2026-11-05 |
| 5 | Whisper 3 Large OpenAI | 4.07 WER % | Whisper Large v3 · fal.ai | — | 2026-11-05 |
| 6 | Whisper 3 Large OpenAI | 4.53 WER % | Whisper Large v3 · Together AI | — | 2026-11-05 |
| 7 | Whisper Large v3 Turbo OpenAI | 4.62 WER % | Whisper Large v3 Turbo · Groq | — | 2026-11-05 |
| 8 | Whisper 3 Large OpenAI | 4.69 WER % | Wizper Large v3 · fal.ai | — | 2026-11-05 |
| 9 | Gemini 2.5 Flash Google | 5.14 WER % | Gemini 2.5 Flash · Google | — | 2026-11-05 |
| 10 | Whisper Large v3 Turbo OpenAI | 5.53 WER % | Whisper Large v3 Turbo, Telnyx · Telnyx · none | — | 2026-11-05 |
| 11 | Whisper 3 Large OpenAI | 10.06 WER % | Whisper Large v3 · Replicate | — | 2026-11-05 |