OpenAI Ships GPT-Live-1.1 for Voice Agents, Cuts Idle Cost to $0.02/Min

What happened
OpenAI has launched GPT-Live-1.1, a new version of its voice model for developers using its API. The model supports full-duplex conversations, meaning it can listen and speak at the same time in multiple languages. OpenAI says this version produces more natural accents and is better at following instructions and handling details. The update also introduces a standby mode that lowers the cost of idle voice agent sessions to $0.02 per minute.
Key details
- Model ID: gpt-live-1.1, available via OpenAI API
- Default context window: 128,000 tokens
- Active session cost: $0.05 per minute; standby (idle) cost: $0.02 per minute
- Standby mode keeps the session open but pauses inference during silence
- No published benchmark scores for GPT-Live-1.1; improvements are self-reported
Why it matters
Teams and companies running voice agents can now reduce their costs during idle periods, as the standby rate is 60% lower than the active rate. Developers gain access to a model that claims better task completion and more natural speech, but must rely on OpenAI's self-reported improvements since no independent benchmarks are available. The change affects how organizations budget for and deploy always-on voice services using OpenAI's platform.
