OpenAI today launched GPT-Live, a new family of AI models optimized for voice commands. The series will power ChatGPT's voice mode and be available to developers via API.
The launch includes two models: GPT-Live-1 for paid ChatGPT plans and GPT-Live-1-mini for the free tier. The key innovation is a full-duplex architecture that allows the AI to interrupt users for real-time tasks like web searches, eliminating previous delays.
GPT-Live routes complex queries to GPT-5.5, OpenAI's current most advanced model. Separately, the company's next-generation GPT-5.6 series, led by model Sol, is set for broad availability this Thursday following U.S. government approval.
OpenAI's new professional services arm, the OpenAI Deployment Company, announced an acquisition of Northslope Inc., a firm specializing in AI applications built on Palantir's software.