OpenAI has introduced GPT-Live, a series of AI models designed to process spoken instructions, scoring 75.5 in pleasantness tests. The models will power ChatGPT's voice mode and are available to developers via an API.
TL;DR
- OpenAI launches GPT-Live voice models with real-time task processing capabilities.
- GPT-Live-1 scores 75.5 in pleasantness tests, outperforming previous voice models.
- Models integrate with GPT-5.5 for advanced tasks and will be available to developers via API.
What happened
OpenAI Group PBC has launched GPT-Live, a family of AI models optimized for spoken instructions. The series includes GPT-Live-1, the more capable model, and GPT-Live-1-mini, a scaled-down version for the free tier of ChatGPT. The models are rolling out across iOS, Android, and web versions of ChatGPT, supporting multiple languages.
GPT-Live features a full-duplex architecture that allows it to interrupt users, search the web, or perform tasks without waiting for the user to finish speaking. This addresses the limitations of ChatGPT's previous voice mode, which could only perform tasks after the user had finished speaking.
Why it matters
For developers, the availability of GPT-Live via an API opens up new possibilities for integrating advanced voice processing capabilities into applications. This could lead to more interactive and efficient user experiences.
For investors, OpenAI's continuous innovation and strategic acquisitions, such as the recent acquisition of Northslope Inc., indicate a strong focus on expanding its professional services and adoption of its LLMs. This could drive further growth and revenue.
Key facts
- GPT-Live-1 scores 75.5 in pleasantness tests, outperforming previous voice models.
- GPT-Live integrates with GPT-5.5 for advanced tasks like web browsing and complex topic explanations.
- Models are available via API for developers to integrate into their applications.
- GPT-Live is rolling out across iOS, Android, and web versions of ChatGPT.
- OpenAI's revenue grew sevenfold last year.
- OpenAI Deployment Company has raised $4 billion in funding.
- GPT-5.6 models, including Sol, Tera, and Luna, will be broadly available on Thursday.
Context
OpenAI's launch of GPT-Live comes ahead of the broad release of GPT-5.6, which has already shown impressive performance in coding benchmarks. The company's focus on voice processing and real-time task management reflects a broader industry trend towards more interactive and efficient AI-driven applications.
The acquisition of Northslope Inc. by the OpenAI Deployment Company underscores OpenAI's commitment to helping organizations adopt its LLMs. This strategic move could further solidify OpenAI's position in the competitive AI landscape.
