OpenAI17:19Feature UpdatesOfficial X
OpenAI Launches GPT-Realtime-2 for Low-Latency Voice Agents
Handles complex voice tasks at low latency, simplifying voice control in CRM workflows.
Key Points
- 1128K context for long conversations
- 2Parallel tool calls and tone control
- 3200-500ms latency for natural flow
- 4Pricing: $32/M input tokens
OpenAI added GPT-Realtime-2 to the Realtime API, enabling GPT-5-level reasoning in voice with 128K context, tool calls, and interruption handling. Translate and Whisper models were also released for comprehensive audio apps. Developers can now build natural conversational agents easily.