Compare
One benchmark. Five comparisons.
Every comparison below uses the same public dataset (CommonVoice 5.1) and the same methodology, replicating Assembly AI's published benchmark, so the numbers are comparable across providers. Prices are each provider's published batch rate against Salad's $0.16 per audio hour volume tier.
| Provider | Accuracy | Price / audio hr | Salad advantage | |
|---|---|---|---|---|
| Salad Transcription API | 95.1% | $0.16 | — | Product → |
| Assembly AI | 93.4% | $0.37 | +1.7% accuracy · 56% less cost | Compare → |
| Azure Batch | 91.2% | $0.36 | +3.9% · 55% less | Compare → |
| Deepgram | 91.0% | $0.26 | +4.1% · 38.4% less | Compare → |
| Google Speech-to-Text | 90.8% | $0.96 | +4.3% · 83% less | Compare → |
| Amazon Transcribe | 89.7% | $1.44 | +5.4% · 89% less | Compare → |
Accuracy = 100% − word error rate on CommonVoice 5.1 English. Competitor prices as published at time of benchmark; verify current list prices with each provider. Full methodology.
Coming
GPU cloud comparisons.
Per-GPU-hour comparisons against major GPU clouds are in preparation and will be published here with the same discipline: public prices, dated, with links.