OpenAI Prepares Major ChatGPT Upgrade With Bidirectional Audio Model

By: Harsh Vardhan

|

Updated on: 24-Jun-2026 01:00 PM

6,302 views

Follow Us:

6,302 views

OpenAI is preparing a significant upgrade to ChatGPT's conversational abilities. The company plans to enhance its AI products, aiming to transform ChatGPT into a super app. These changes include improvements to OpenAI's Codex coding tool and agentic AI tools that perform tasks for users.

Key Highlights

• OpenAI is developing GPT-Bidi-1, a bidirectional audio model for ChatGPT.
• GPT-Bidi-1 can speak, hear, and listen simultaneously for more natural conversations.
• The model remembers earlier conversation parts and handles interruptions smoothly.
• Early rollout has begun for a small group of ChatGPT app users.

New Bidirectional Audio Model

Reports indicate that OpenAI is developing a new audio model called GPT-Bidi-1. The name "Bidi" stands for bidirectional, reflecting its ability to speak, hear, and listen at the same time. TestingCatalog first identified references to GPT-Bidi-1 last week. Internal code describes the model as a "major leap in intelligence" and "the next generation of Voice."

GPT-Bidi-1 aims to make conversations more natural. It can respond with brief acknowledgements, such as "okay," when users pause or slow down. This feature allows the assistant to maintain the flow of conversation without interrupting. The model also handles interruptions more effectively. For example, if a user asks it to count from one to ten and then interrupts to reverse the count, GPT-Bidi-1 can adjust immediately.

Improvements in Conversation Flow

One of the most notable changes is the model's ability to remember earlier parts of a conversation. GPT-Bidi-1 can maintain context over longer discussions, responding based on previous exchanges. Reports also state that the model no longer jumps into conversations during long pauses, addressing a common frustration with ChatGPT's current voice mode.

Users may find GPT-Bidi-1 in ChatGPT's model selector, alongside standard and advanced voice options. When selected, the voice bubble reportedly turns yellow. These details come from code references, user interface sightings, and early user tests shared by multiple sources. However, OpenAI has not released a technical whitepaper or engineering blog about the model's capabilities.

Rollout and Future Implications

TestingCatalog reports that GPT-Bidi-1 has started rolling out to a small group of ChatGPT app users. This suggests a broader release could follow soon. The new model may help OpenAI close the gap between its advanced text models and older voice technology.

OpenAI is betting that speech will become a primary way people interact with AI. If GPT-Bidi-1 performs as reported, it could make conversations with ChatGPT feel more like speaking to a person. The assistant would be able to listen, understand, and respond in real time, improving the user experience.

These developments highlight OpenAI's ongoing efforts to advance conversational AI and enhance user interaction through improved voice technology.