OpenAI, on Thursday, unveiled three audio models for its developer platform, with the goal of improving the conversational abilities of voice-based software agents and enabling them to perform tasks in real time.
The introduction of the application programming interface (API) signifies a shift for the ChatGPT creator from simple transcription and chat functionalities to developing agents capable of listening, translating, and acting during live interactions.
The newly launched models are GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper.
OpenAI has announced that these models are available for testing in its developer playground.
Three new audio models by OpenAI
GPT-Realtime-2 is specifically engineered to handle more complex requests, use tools, manage interruptions, and sustain context throughout extended voice sessions.
On the other hand, GPT-Realtime-Translate facilitates translation from over 70 languages into 13 target languages, making it particularly useful for customer support, educational purposes, and various other applications.
The third model, GPT-Realtime-Whisper offers real-time speech-to-text capabilities, enabling the generation of captions, meeting notes, and workflow updates as a speaker communicates.
Among the customers currently testing these models are the online real estate marketplace Zillow, the online travel agency Priceline, and the European telecommunications company Deutsche Telekom.
OpenAI launches ChatGPT 5.5
The new launch comes after OpenAI announced the introduction of its latest artificial intelligence (AI) model, GPT-5.5.
The release of GPT 5.5 occurs less than two months following OpenAI’s release of GPT 5.4, which is the latest indication of the rapid development pace taking the AI industry forward.
“What is particularly remarkable about this model is its improved capability to perform tasks with minimal guidance,” stated OpenAI President Greg Brockman during a press briefing.
Read more: OpenAI launches ChatGPT 5.5, taking on competitors Google and Anthropic
