OpenAI's GPT‑Bidi‑1 voice model enters testing phase

OpenAI is developing a new bidirectional audio model for ChatGPT, referred to as GPT‑Bidi‑1 (also reported as GPT‑BiDi‑1). The model is designed to listen and speak at the same time, allowing real‑time interaction, handling interruptions without restarting the response, and acknowledging pauses with short acknowledgments such as “okay.” Code references and UI elements indicating a selectable “yellow voice bubble” were discovered in the ChatGPT mobile app in mid‑June 2026, suggesting the feature is under internal testing and has already been rolled out to a limited group of users. The upgrade aims to make voice conversations feel more natural and fluid, improving context retention and reducing the awkward pauses that current voice modes exhibit. No official public announcement has been made yet.