Photo: OpenAI
ChatGPT will now respond to messages and interact with all participants in shared group conversations.
OpenAI has announced the global rollout of group chats in ChatGPT after a successful week-long test in Japan, New Zealand, South Korea, and Taiwan. The feature will become available within the next few days to all users — Free, Go, Plus, and Pro.
According to OpenAI, group chats allow up to 20 people to participate in a single conversation, where everyone can send prompts to ChatGPT, react to messages, share images and files, and take part in real-time discussions. Users can join a chat via a personal invite link with an existing account or by creating a new one.
In this new format, ChatGPT runs on GPT-5.1 Auto — a system that automatically selects the most suitable model for each reply based on the request and the participant’s subscription tier. OpenAI says the model also features improved “social” behavior: it follows the flow of conversation and responds only when appropriate, unless directly mentioned by name.
Before starting a group chat, users will be prompted to set up a profile with a name and photo. All shared conversations will appear in a dedicated section on the sidebar.
The settings menu allows users to manage participants, change the chat topic, and toggle notifications. Chat creators can also customize how ChatGPT behaves in each group by assigning it a specific tone or role.
“Group chats bring people and ChatGPT together in a single conversation. For example, if you’re planning a weekend trip with friends, you can create a group chat where ChatGPT helps compare destinations, build an itinerary, and keep everyone involved,” the company explains.
OpenAI emphasizes that group chats remain fully separate from a user’s private conversations and do not interact with the model’s personal memory. ChatGPT does not form new memories from group discussions, and participants can leave the chat or remove others at any time — except for the original creator.
Additional protections apply for minors. If a user under 18 is present, the content level is automatically restricted for everyone. Parents and guardians can also disable group chats entirely via parental controls.
The rollout comes alongside the launch of the new GPT-5.1 model, which introduces two modes: the fast Instant mode for everyday use and the analytical Thinking mode for complex tasks. A key feature of the update is deeper personalization, allowing the assistant to adapt its tone and communication style for more natural, human-like conversations.