OpenAI has begun rolling out GPT-Live, a caller procreation of sound models that powerfulness ChatGPT Voice. During a conversation, GPT-Live tin manus a mobility to its latest frontier model, specified arsenic GPT-5.5, for heavier reasoning aliases a web search.
What Launched
GPT-Live-1 becomes the default exemplary powering ChatGPT Voice for Go, Plus, and Pro users, and GPT-Live-1 mini becomes the default forusers.
OpenAI describes the models arsenic full-duplex, meaning they tin return successful audio and nutrient reside astatine the aforesaid time, interrupt less, and hold erstwhile a personification pauses.
Search and Visual Answers Move Into the Conversation
When a spoken mobility needs deeper reasoning aliases existent information, GPT-Live tin manus it to a frontier model, GPT-5.5 astatine launch, and bring the reply backmost into the conversation.
ChatGPT Voice tin besides show ocular cards for topics for illustration weather, stocks, and sports, and OpenAI said Voice continues to support search, memory, images, and record uploads. Users tin group a reasoning level, pinch Instant for accelerated replies and Medium aliases High for much progressive questions. OpenAI said Instant replies and the mini exemplary tally connected GPT-5.5 Instant, while Medium and High usage GPT-5.5 Thinking.
How OpenAI Measured It
In OpenAI’s ain head-to-head evaluations, GPT-Live-1 and GPT-Live-1 mini were preferred complete its Advanced Voice Mode successful conversations lasting 5 to 10 minutes. OpenAI said the evaluations looked astatine wide preference, turn-taking, interruptions, conversational flow, and really earthy the interactions felt. The institution said much than 150 cardinal group talk to ChatGPT each week utilizing features for illustration Voice and Dictation.
What It Doesn’t Do Yet
At launch, GPT-Live does not support sound pinch video aliases surface sharing successful ChatGPT. OpenAI said it is moving to adhd those capabilities, and that its earlier Standard and Advanced Voice Modes enactment disposable wherever video and surface sharing are supported.
Why This Matters
Now, erstwhile you inquire a mobility aloud, you mightiness spot an reply straight successful the sound flow, sometimes accompanied by a ocular paper connected the screen. This provides ChatGPT pinch different measurement to coming an reply without visiting a root site. Since the reasoning and hunt process happens down the scenes connected GPT-5.5, what shows up location depends connected really the exemplary retrieves accusation and cites its sources.
Looking Ahead
OpenAI said the rollout is opening globally, and that video and surface sharing are not successful this merchandise but are being worked on.
OpenAI’s station doesn’t opportunity really GPT-Live handles citations erstwhile it answers a spoken mobility from a GPT-5.5 web search. ChatGPT’s matter answers show root links adjacent to the response. Whether a spoken reply names its sources, shows them connected screen, aliases leaves them retired is the item to watch. That’s what decides whether a hunt wrong a sound speech tin still nonstop a scholar to your site.
Featured Image: OpenAI
English (US) · 
Indonesian (ID) ·