OpenAI launched GPT-Live-1 and GPT-Live-1 mini, full-duplex conversational voice models that can simultaneously listen and speak, enabling natural interruption, live translation, and extended multi-turn sessions. The architecture replaces the prior three-stage pipeline (speech-to-text, LLM, text-to-speech) with a unified model that routes to backend text models such as GPT-5.5 for search, reasoning, and agentic tasks. GPT-Live-1 mini becomes the default for ChatGPT's Advanced Voice Mode, with GPT-Live-1 reserved for paid tiers, and OpenAI cited 150 million users engaging with voice and dictation features. For financial institutions, the tiered access structure and the positioning of voice as a primary interface to agentic workflows are relevant signals for both consumer monetization and enterprise deployment surface area.
324 other tracked developments touching the same claims.