OpenAI announced that it has begun offering GPT-Live-1 via API as a powerful voice model for developers to build voice-enabled applications and business workflows. This deployment brings features introduced in ChatGPT to the API, allowing for the simultaneous listening and speaking of audio.
Model ReleasesPricing & LimitsOpenAIGPT-Live-1
OpenAI Launches Voice Model GPT-Live-1 for API
This article is a translation. Read the Japanese original
While conventional voice agents sequentially linked the processes of speech recognition, inference, and speech synthesis, GPT-Live-1 handles these within a single model. This makes it easier to maintain the rhythm and context of a conversation. Furthermore, it is designed to respond in real-time to user interruptions and backchanneling.
Developers can freely combine backend inference models according to their specific needs. For example, a lightweight model can be paired for routine tasks such as scheduling, while a model capable of advanced inference can be used for complex customer support. The pricing is reported to be $0.05 per minute for the frontend voice layer.
Sources
- Build more natural voice experiences with GPT‑Live‑1 in the API （OpenAI News, 2026-09-10）