OpenAI is on the verge of a major update to its flagship product, ChatGPT. Insider information and code analysis of the platform point to the imminent launch of a fundamentally new voice interface that could radically change how humans interact with artificial intelligence.
Technological breakthrough: what is GPT Bidi 1
According to industry experts and information published in the Testing Catalog, a limited group of users has already gained access to the test version of the new model. The application code features the name GPT Bidi 1. The acronym Bidi stands for bidirectional design.
This architectural feature allows the assistant not just to react to commands, but to speak, listen, and analyze the user's speech simultaneously. OpenAI developers are betting specifically on the voice interface, considering it the primary way to access AI in the future. The new model aims to bridge the gap between advanced text architectures and the outdated voice layer.
How it works: naturalness and instant reaction
The key difference of the new version is its ability to conduct a dialogue as naturally as possible. Unlike previous versions, where the AI often lost the thread of the conversation or interrupted the user inappropriately, GPT Bidi 1 demonstrates a high degree of contextual understanding.
The system has learned to insert short confirmation phrases, such as "okay," when the interlocutor pauses, without interrupting the general flow of the conversation. Additionally, the algorithm no longer jumps into the conversation prematurely during long pauses, making communication more comfortable.
Special attention has been paid to instant task switching. The assistant is capable of changing its action algorithm "on the fly." For example, if you ask the model to count to ten and then ask it to count backwards during the process, the system will instantly adapt to the new instructions without resetting the context.
Visual changes and context
In the application settings, the new tool has already appeared in the model selector alongside standard and advanced options. The visual marker for activating GPT Bidi 1 will be a change in the color of the voice interface graphic bubble: it will turn yellow instead of the usual blue.
One of the most important advantages of the new architecture is the ability to maintain the logic of a long conversation within a single context. In previous versions, prolonged communication often led to the loss of early fragments of the dialogue, which was considered a serious flaw of the system. GPT Bidi 1 solves this problem, ensuring the integrity of the conversation.
Official release
At this time, OpenAI has not officially announced the release of the GPT Bidi 1 model and has not provided detailed specifications regarding the next version of GPT 5.6. However, the active deployment of the interface among the first testers indicates that the product is in the final stages of preparation for a global release.
