AssemblyAI vs Deepgram
AssemblyAI Universal-3 Pro vs Deepgram Nova-3 for transcription. Price: $3.5 per 1000 min vs $4.3 per 1000 min — edge AssemblyAI WER: 3.1% vs 5.2% — edge AssemblyAI Langs: 99 vs 50 — edge AssemblyAI Computed from public benchmarks with dated sources; updated 2026-08-20.
Head to head
| Axis | AssemblyAI Universal-3 Pro | Deepgram Nova-3 | Edge |
|---|---|---|---|
| Price | $3.5 per 1000 min | $4.3 per 1000 min | AssemblyAI |
| WER | 3.1% | 5.2% | AssemblyAI |
| Langs | 99 | 50 | AssemblyAI |
| Latency | 150 ms | 300 ms | AssemblyAI |
| diarization | ✓ | ✓ | — |
| timestamps | ✓ | ✓ | — |
| vocab | ✓ | ✓ | — |
Strengths & caveats
AssemblyAICheapest of the directly-comparable dedicated APIs ($3.50/1000 min) with the widest language coverage (99) and full diarization, timestamps and custom vocabulary; ~150 ms P50 streaming. Accuracy (3.1% WER) trails the leaders; the cheaper legacy Universal-2 tier is less accurate still. AssemblyAI now lists Universal-3.5 Pro as its flagship at the same $3.50/1000 min, but no independent leaderboard score exists for it yet, so the WER here still refers to Universal-3 Pro.DeepgramStreaming-focused with per-second billing and full diarization, timestamps and keyterm support; markets sub-300 ms latency. Lowest accuracy of the group on the leaderboard (5.2% WER); the latency figure is self-reported, not an independent benchmark. Deepgram now positions a separate, dearer Flux model ($0.0077/min list) as its ultra-low-latency option for voice agents.
Sources
- Artificial Analysis — Speech to Text2026-08-20
- Open ASR Leaderboard2026-06-19