OpenAI has released new voice models, including GPT-Live and GPT-Live-1 mini, which are full-duplex models that can listen and speak simultaneously. These models delegate deeper reasoning to GPT-5.5, and are intended to facilitate more natural live conversations. The release appears to be part of OpenAI's efforts to advance voice-based interactions, with some outlets suggesting it may be a precursor to a broader GPT-5.6 release.