How OpenAI delivers low-latency voice AI at scale
OpenAI has released three new real-time voice models: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. These models offer enhanced reasoning capabilities, live speech translation for over 70 languages, and low-latency transcription. GPT-Realtime-2, in particular, is described as having "GPT-5-class reasoning" and features a significantly expanded context window of 128K tokens, alongside improved handling of interruptions and tool usage. AI
IMPACT Enhances real-time voice agent capabilities with improved reasoning, translation, and transcription, potentially accelerating adoption of voice-first interfaces.