OpenAI introduces a new AI voice model that can think and speak in more than 70 languages

OpenAI has introduced three new audio models for its Realtime API, marking a major step forward in the development of voice-powered artificial intelligence applications.

The new models — GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper — are designed to help developers build AI systems capable of understanding, responding, translating, and transcribing speech in real-time.

Leading the launch is GPT-Realtime-2, a voice model powered by GPT-5 level reasoning capabilities. According to OpenAI, the model can manage more complex conversations while maintaining context in longer interactions.

The company says the model supports simultaneous use of the tools, allowing it to perform tasks such as checking a calendar or retrieving information while giving the user verbal updates about what it is doing. It also features a 128K token context window, allowing for longer and more coherent conversations.

READ ALSO: AI giant OpenAI raises record $122 billion in funding, eyes IPO

OpenAI also launched GPT-Realtime-Translate, a real-time translation model capable of understanding more than 70 input languages ​​and providing responses in 13 output languages. The company says the system is designed to keep up with the natural flow of conversation rather than offering delayed translations.

The third model, GPT-Realtime-Whisper, focuses on live transcription by converting spoken words into text as the conversation occurs. This capability is aimed at developers building voice assistants, customer service tools, meeting software, and multilingual communications platforms.

“Together, the models we launched move real-time audio from simple call and response to a voice interface that can actually do the work: listen, reason, translate, transcribe, and take action as the conversation progresses,” OpenAI said in a statement.

The launch reflects increasing competition in the AI ​​industry to develop more natural and capable voice interfaces that go beyond basic chatbot interactions.