OpenAI has unveiled GPT-Live, a new full-duplex model designed to enhance ChatGPT Voice with real-time, human-like conversational AI. The model, which allows simultaneous listening and speaking, marks a significant advancement in natural language interaction.
Key Features of GPT-Live
GPT-Live mimics human conversation by incorporating filler phrases such as 'mhmm' or 'yeah', along with quick responses and natural pauses. This capability is powered by GPT-5.5, enabling more fluid and engaging dialogues. According to OpenAI, the model aims to make interactions with ChatGPT feel as natural as speaking with another person.
Dual Model Rollout
OpenAI is releasing two versions of the model: GPT-Live-1 and GPT-Live-1 mini. The standard GPT-Live-1 will be the default for paid users, while GPT-Live-1 mini is designed for free users. Both versions are available across mobile and web platforms, ensuring broad accessibility.
Custom Silicon Development
In parallel, OpenAI is collaborating with Broadcom to develop a custom chipset codenamed 'Jalapeño'. This initiative aims to improve efficiency and performance-per-watt for AI workloads, reducing reliance on third-party hardware. The move reflects a broader industry trend, with rivals like Anthropic also exploring custom silicon partnerships with Samsung to optimize AI data centres.
