GPT-Live-1.1 is rolling out in ChatGPT Voice with more natural accents and new voices, alongside the smaller GPT-Live-1.1 Quip. OpenAI announced the update on October 15, 2026. For users, the immediate question is which voice experience their account offers; for developers, it is whether the announcement changes the model they can deploy through the API.
The distinction matters: a ChatGPT rollout does not automatically establish a new API model. Here is what the official sources confirm, and how we would evaluate the update.
What changed in GPT-Live-1.1?
According to OpenAI’s October 15 release notes, the update improves accents across languages and varies the brief phrases Live uses while searching. New voice options cover French, German, Hindi, Japanese, Korean and Mexican Spanish, with Bubbie added for English. OpenAI says the available voices can differ by account and region.
These are conversational improvements. The release entry does not provide a measured reduction in search time or a benchmark demonstrating better factual accuracy. More varied progress speech can make waiting feel less repetitive, but it should not be treated as evidence that the underlying task completes faster.
How to try the new voices
Open Settings → Voice to select the Live experience, then use the Voice selector to choose an available voice. The ChatGPT Voice help article identifies GPT-Live-1.1 and Quip as the models behind Live, with access depending on the plan. Workspace settings and other account conditions can also affect availability.
The documented language options include Aube and Rivage for French, Sonna and Miko for German, Leher and Neer for Hindi, Himari and Rindō for Japanese, Haetsal and Hanul for Korean, and Tinta and Arrullo for Mexican Spanish. A voice missing from one account is not enough evidence that the entire release is unavailable.
For an initial comparison, use the same short task with your previous voice and an available new option. Include a name, a correction and a sentence in the language you use most. Judge whether the conversation is easier to follow and whether the final answer preserves your correction. A pleasing voice is useful, but successful communication also requires the right information.
Live, Advanced and Standard serve different needs
OpenAI documents Live as able to listen while speaking. It does not support video or screen sharing; eligible mobile users can use Advanced for those capabilities. Standard transcribes speech before responding.
Choose the experience around the task. Talking through a written proposal and showing a changing screen have different input requirements. Before replacing a familiar workflow, check that its required inputs remain available in the selected mode. Our recommendation is to compare a complete task, including any visual context it needs, rather than judge only the opening exchange.
Does GPT-Live-1.1 change the developer API?
As checked on October 16, the official API model page still lists gpt-live-1 as its model ID and snapshot. The ChatGPT release does not establish an available gpt-live-1.1 API identifier. Developers should verify supported IDs before changing production configuration.
There are separate API updates nearby: the API changelog dates native Decisions integration to October 15 and prompt-to-voice plus nine additional voices to October 16. Those entries concern GPT-Live 1. They should be evaluated as their own API features, rather than relabeled as a confirmed 1.1 API launch.
What teams should evaluate next
Our editorial assessment is that this release is most relevant to teams whose spoken workflows depend on language comfort and understandable progress updates. A useful evaluation would measure how often users repeat themselves, whether corrections survive to the final result, and whether people can tell when a search is still running.
For example, ask the assistant to research a supplier, correct the supplier name halfway through, then inspect the final answer. Record both conversational friction and task correctness. Repeat with representative speakers and ordinary background conditions. This is a proposed evaluation, not a claim that we have tested the release.
GPT-Live-1.1 gives ChatGPT users new conversation options to assess. The production decision should follow evidence from the actual workflow: clearer speech, fewer misunderstandings and a correct completed task.