What happened
OpenAI is reportedly preparing GPT-Bidi-1, a new bidirectional voice model for ChatGPT, enabling simultaneous listening and speaking. This model allows ChatGPT to respond with acknowledgements like "okay" during user pauses, handle mid-sentence interruptions by adjusting immediately, and maintain context across long conversations. TestingCatalog first spotted references to GPT-Bidi-1, described internally as a "major leap in intelligence" and "next generation of Voice," with early rollout reportedly underway for a small group of ChatGPT app users.
Why it matters
Conversational AI interactions will become significantly more natural, reducing friction for users and expanding voice-led application possibilities. For product managers and UX designers, this mechanism shifts voice interface design from sequential command-response to fluid, interleaved dialogue, mirroring human conversation. This follows OpenAI's earlier launch of real-time voice models, indicating a strategic prioritisation of speech as a primary AI interaction method. Procurement teams should anticipate increased demand for infrastructure supporting real-time, low-latency audio processing.
