OpenAI Launches GPT-Realtime-2, Bringing GPT-5-Class Reasoning to Voice Agents
OpenAI released GPT-Realtime-2 in the Realtime API, adding GPT-5-class reasoning to voice agents alongside a 70-language translation model.
OpenAI has released GPT-Realtime-2 in its Realtime API, describing it as the company's most intelligent voice model to date. The model brings GPT-5-class reasoning to voice agents, enabling them to handle complex, multi-turn conversations in real time. A companion model, GPT-Realtime-Translate, ships alongside it with support for streaming translation across more than 70 languages.
For developers building voice-first applications, the launch represents a meaningful capability step: agents can now listen, reason, take action, and handle interruptions within live conversations rather than relying on a separate transcription-then-reasoning pipeline. OpenAI positioned the release as API-first and production-ready, noting that voice updates for ChatGPT are still separate and forthcoming. Both models are available in the Realtime API now.