Home › Voice agents › OpenAI Realtime API
OpenAI Realtime API 7.4/10
Speech-to-speech model API for low-latency voice agents.
| Category | Voice agents |
|---|---|
| Pricing model | usage-based |
| Starting price | $32/1M audio input tokens, $64/1M output (gpt-realtime-2.1); mini is cheaper |
| Open source | No |
| Best for | Builders wanting a single speech-to-speech model |
| Limitations | Token-based audio pricing is hard to map to minutes; OpenAI models only. |
| Last checked |
What it can do
- Native speech-to-speech
- Tool calling
- SIP and WebRTC
Score breakdown
| autonomy | 7/10 |
|---|---|
| reliability | 8/10 |
| ease | 7/10 |
| value | 7/10 |
| ecosystem | 8/10 |
Natural low-latency model; DIY around it.
For agencies
Use cases: client support botsvoice receptionists
| White-label | Unknown |
|---|---|
| Multi-client / workspaces | Unknown |
Cost per client, setup time and what broke: not tested yet. This section fills in once OpenAI Realtime API gets a Henchwork test.