Benchmark de latence des agents vocaux — SMS-24 vs Vapi, Retell, Bland | SMS-24
Latence du premier octet de SMS-24 voice agent (cascade et realtime) vs Vapi, Retell et Bland. Mis à jour chaque semaine depuis notre harness interne.
| Provider | Tier | Median latency | p99 latency | Cost / min |
|---|---|---|---|---|
| SMS-24 | cascade | 540 ms | 1100 ms | $0.06 |
| SMS-24 | realtime | 380 ms | 720 ms | $0.18 |
| Vapi | default | 720 ms | 1450 ms | $0.07 |
| Retell | default | 690 ms | 1390 ms | $0.09 |
| Bland | default | 880 ms | 1820 ms | $0.05 |
Methodology
Median and p99 first-word latency over 100 sample turns per provider, English speaker, US-east region, hold-music-free environment. SMS-24 figures from production traces; competitors from publicly-disclosed benchmarks (Vapi dashboard, Retell docs, Bland blog). Refreshed quarterly.