OpenAI’s October 15 GPT-Live-1.1 system card explicitly names Dots alongside Codex among the experiences using its new default voice model for Plus and higher tiers. The company describes GPT-Live-1.1 and the smaller Quip variant as successors to its existing Live models. Both support continuous listening and responding, including handling interruptions and changes in conversational pace. This is a documented voice-model update, rather than evidence that Dots has gained a new autonomous task capability.
The accompanying ChatGPT release notes say rollout is under way and describe more natural accents, additional short spoken cues during search, and new voice options in French, German, Hindi, Japanese, Korean and Mexican Spanish, plus Bubbie in English. OpenAI cautions that the voices shown can vary by region and account. The system card connects the model to Dots, but that does not establish that every voice option has reached every Dots account or client. We have not tested the rollout or compared its conversational quality ourselves.
Editorial take: For people who brief an agent aloud, the useful question is whether the conversation remains easy to interrupt and clarify while work is being delegated. A more natural voice can improve that interaction, but it is still worth checking the written task and result before relying on them. Treat the release as an interface improvement whose availability needs checking in your account, without inferring broader permissions or better task outcomes from the sound of the conversation.