OpenAI Launches Advanced Realtime Voice AI Models
OpenAI has introduced three new realtime voice models that bring advanced reasoning, live translation, and streaming transcription to voice applications. The new models include GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. Consequently, developers can build AI assistants that listen, reason, translate, and respond naturally during live conversations.
The release expands OpenAI’s Realtime API and moves voice AI beyond simple command-response interactions. Moreover, the models are designed to support enterprise applications, customer service, education, and multilingual communication.
New Models Enhance Voice Intelligence
GPT-Realtime-2 is OpenAI’s first voice model with GPT-5-class reasoning capabilities. Therefore, it can manage complex requests, maintain context during long conversations, handle interruptions, and call external tools while interacting with users. Developers can also adjust reasoning levels based on application requirements.
OpenAI also launched GPT-Realtime-Translate, which supports speech translation from more than 70 input languages into 13 output languages. As a result, conversations continue naturally without noticeable pauses while preserving meaning and context. The company expects the model to benefit customer support, travel, education, and international business applications.
Meanwhile, GPT-Realtime-Whisper delivers low-latency speech-to-text transcription. In addition, it generates live captions, meeting notes, and workflow updates as speakers continue talking, making it suitable for productivity and accessibility use cases.
Enterprise Adoption Begins
Several organizations have already begun evaluating the new models. These include Zillow, Priceline, Deutsche Telekom, Vimeo, and BolnaAI, which are testing the technology across customer engagement, multilingual communication, and AI-powered services. Furthermore, OpenAI says the models integrate directly into its Playground and Realtime API, allowing developers to begin building immediately.
OpenAI has also introduced usage-based pricing. GPT-Realtime-2 is billed by audio tokens, while GPT-Realtime-Translate and GPT-Realtime-Whisper use per-minute pricing. Consequently, organizations can choose services based on their application needs and deployment scale.
Voice AI Enters a New Phase
Demand for intelligent voice interfaces continues to grow across industries. Therefore, technology companies are investing in models that can understand context, perform actions, and support multilingual conversations in real time.
With these new models, OpenAI strengthens its position in conversational AI while giving developers more powerful tools for building production-ready voice applications. As realtime AI adoption accelerates, advanced voice reasoning, translation, and transcription are expected to become core capabilities across enterprise and consumer platforms.
