Mangaluru: OpenAI appears to be preparing a major upgrade to ChatGPT‘s conversational capabilities, with reports suggesting the company is testing a new bidirectional audio model named GPT-Bidi-1 that could make interactions with artificial intelligence feel more natural and human-like.

The reported development is said to be part of OpenAI’s broader ambition to transform ChatGPT into a “super app”, with significant improvements expected across its AI products, including coding and agentic AI tools capable of performing tasks on behalf of users.

According to reports, GPT-Bidi-1 has been designed to enhance voice interactions by enabling simultaneous speaking and listening, potentially creating a more fluid and intuitive conversation experience.

Bidirectional model aims to improve conversations

The name “Bidi” reportedly stands for “bidirectional”, referring to a system that allows the AI assistant to speak, hear and process information at the same time.

The new model was first spotted by TestingCatalog, which reported references to GPT-Bidi-1 in internal code. According to the report, the model has been described internally as a “major leap in intelligence” and “the next generation of Voice”.

Unlike conventional voice assistants that often wait for users to finish speaking before responding, GPT-Bidi-1 is said to be capable of providing subtle acknowledgements during conversations. For example, it may respond with phrases such as “okay” when a user pauses or slows down, creating a more engaging and natural interaction without interrupting the speaker.

Better handling of interruptions

One of the key features reportedly being tested is the model’s ability to manage interruptions smoothly.

Reports suggest that if a user asks the assistant to count from one to ten and then changes the instruction midway, asking it to reverse the count instead, the model can adapt instantly and continue the interaction seamlessly.

This marks a significant improvement over many existing voice assistants, which often struggle to adjust when users interrupt or change instructions during a conversation.

The ability to understand and respond dynamically could make GPT-Bidi-1 particularly useful in real-time conversations, customer support applications and voice-based productivity tools.

Improved memory and contextual understanding

Another reported enhancement is the model’s ability to retain context during extended conversations.

Current voice assistants can sometimes lose track of earlier parts of a discussion or respond in ways that feel disconnected from the ongoing conversation. GPT-Bidi-1, however, is said to remember previous exchanges and use that information to generate more coherent and contextually relevant responses.

Reports also indicate that the model no longer jumps into conversations during long pauses, a behaviour that has occasionally frustrated users of ChatGPT’s current voice mode.

These changes could significantly improve the overall user experience by making interactions feel less robotic and more akin to speaking with another person.

Possible rollout already underway

According to recent reports from TestingCatalog, GPT-Bidi-1 may have already started rolling out to a limited group of ChatGPT users through the mobile application.

Some reports suggest that users selecting the model may see a yellow-coloured voice bubble, and the feature could eventually appear in ChatGPT’s model selector alongside the standard and advanced voice options.

However, OpenAI has not officially confirmed the existence of GPT-Bidi-1, nor has it published a technical paper or engineering blog detailing the model’s capabilities.

Voice seen as the future of AI interaction

The reported development comes as technology companies increasingly view voice interactions as the future of human-AI communication. Industry experts believe many users may eventually prefer speaking to AI assistants instead of typing queries.

If the reports prove accurate, GPT-Bidi-1 could help bridge the gap between OpenAI’s advanced text models and its existing voice technology, making conversations with AI assistants more responsive, intuitive and human-like.

While an official announcement is still awaited, the early reports have generated considerable excitement among technology enthusiasts eager to see the next evolution of AI-powered voice interactions.