OpenAI is preparing GPT-Bidi-1, a bidirectional voice model that lets ChatGPT listen and speak at the same time, handle mid-conversation interruptions naturally, and translate speech in real time. The upgrade is expected to ship alongside GPT-5.6 this Thursday, June 26, marking the most significant leap in ChatGPT voice since Advanced Voice Mode launched in 2024. OpenAI has not made a formal announcement, but internal staging logs and leaked demos surfaced this week.
What Bidi1 Does — The Three Capabilities That Change Everything
Today’s ChatGPT Voice Mode works one direction at a time: the model speaks, you wait, you respond. GPT-Bidi-1 breaks that pattern. The system can listen and speak simultaneously, so when you interrupt mid-sentence it actually stops, acknowledges the change, and continues with your correction incorporated — no freeze, no restart.
Testing Catalog, which first surfaced the model in OpenAI’s internal logs, shared a demo where a user asked Bidi1 to count to ten. Mid-count, the user interrupted with a request to reverse the sequence. The model replied “okay” and immediately switched to counting down. That responsiveness does not exist in any current AI voice product. Beyond interruption handling, Bidi1 adds real-time voice translation — speak in English, the model responds in your target language as you talk — and three intelligence tiers (High, Medium, Instant) so users can trade reasoning depth for lower latency. According to recent reporting, the voice API access is also planned for developers.
Who Gets It First — ChatGPT Plans and the Launch Window
OpenAI has not confirmed which subscription tiers receive Bidi1 at launch. Based on how OpenAI has rolled out previous model upgrades, ChatGPT Pro subscribers typically get first access to new model tiers, with Plus following within weeks. The Instant tier hints at a free-tier path — though likely rate-limited. Polymarket prediction markets place a 90% probability on a GPT-5.6 family launch before June 28.
Why Now — The AI Voice Race Is Moving Fast
Voice is the next competitive frontier. Google Gemini Live handles natural conversation with interruption support. Apple’s Gemini-powered Siri in iOS 27 promises tighter on-device integration than ever. OpenAI’s response is Bidi1: not just a model that responds to voice, but one that genuinely listens while it speaks. If Thursday’s launch delivers on the demos, ChatGPT voice will leap ahead of every current alternative — including Claude, which has no equivalent voice product.
The bigger picture: the battle between ChatGPT, Claude, and Gemini in 2026 is increasingly being fought on product quality, not model benchmarks. A voice mode that feels human is a retention tool. OpenAI knows this.
Our Take
Bidi1 fixes the biggest frustration with AI voice — the robotic wait-and-respond rhythm that makes it feel like a phone tree. Natural interruption handling and real-time translation are not incremental features; they are the floor for a voice assistant anyone would actually use hands-free. The real test is latency under production load: demos are smooth, but OpenAI’s reliability record in June has been shaky. Watch Thursday. If it ships cleanly, this is the most meaningful ChatGPT upgrade of the year.
Frequently Asked Questions
What is GPT-Bidi-1?
GPT-Bidi-1 (Bidi1) is OpenAI’s upcoming bidirectional voice model for ChatGPT. Unlike the current Voice Mode, it can listen and speak simultaneously, allowing natural mid-conversation interruptions without freezing or restarting.
When does the new ChatGPT voice mode launch?
Bidi1 is expected to ship alongside GPT-5.6 on or around Thursday, June 26, 2026. OpenAI has not made a formal announcement, but internal logs and prediction markets both point to a late-June launch.
Will Bidi1 be available via API?
Based on leaked details, an API for Bidi1 is planned for developers, though the exact rollout timeline and pricing have not been confirmed by OpenAI.
