Live source snapshot
Voice Agent Latency Benchmark
Measures the silence callers experience between finishing a turn and hearing a voice-agent platform begin its reply.
Source snapshots refreshed
Five measurement lanesVOICE / S2S
Reasoning, task completion, conversation dynamics, experience, and latency stay separate.
Benchmark profile
- Scope
- Real synchronized phone calls with published recordings, per-turn measurements, discards, and configuration receipts
- Primary metric
- Median and p95 Time To First Audio Byte in milliseconds (lower is better)
- Owner
- OpenBenchmarks Labs
- Available evidence
- Machine-readable public API and reproducible raw-call artifacts
What it measures
Latency
How long do responses, tool calls, and full tasks take?
Available results
| Rank | Platform | Median | p95 | p95 / median | Usable turns |
|---|---|---|---|---|---|
| 1 | Telnyx | 1,296ms | 1,856ms | 1.43× | 419 / 432 |
| 2 | ElevenLabs | 1,424ms | 1,768ms | 1.24× | 429 / 432 |
| 3 | Bland AI | 1,520ms | 2,247.6ms | 1.48× | 429 / 432 |
| 4 | Vapi | 1,558ms | 2,007.8ms | 1.29× | 382 / 432 |
| 5 | Retell AI | 1,740ms | 2,258.6ms | 1.30× | 419 / 430 |
Interpretation limit
This board measures platform-level phone latency, not answer quality or raw speech-model latency. The carrier path and each platform’s deployed stack remain inside the measurement.