OpenAI makes GPT-Live-1.1 available in its API
A post dated November 19, 2026, from OpenAI Developers announced that GPT-Live-1.1 was available in the API; OpenAI says it has powered voice in ChatGPT Plus and Pro since October 15. OpenAI says developers can use it to power apps with voice agents that understand details, follow instructions and hand off complex work to a backend model.
Key takeaways · 4
- 01
GPT-Live-1.1 is now available in OpenAI’s API for powering apps.
- 02
The model can take audio and text as input and return audio and text as output.
- 03
OpenAI says developers can choose a backend model for complex workflows, with delegation happening without breaking the conversation.
- 04
Budget for the live session and any backend model or tool use separately.
From ChatGPT voice to the API
OpenAI Developers announced the API availability on November 19, 2026.[1] OpenAI says GPT-Live-1.1 has powered voice in ChatGPT Plus and Pro since October 15.[2] The announcement now positions the model for developers building their own apps.[2] OpenAI describes its voice agents as able to reliably understand precise details such as letters and numbers, and to follow instructions.[2]
Audio in, audio out
OpenAI lists audio and text as both input and output modalities for GPT-Live-1.1.[3] The model can listen while it speaks, according to OpenAI’s developer documentation.[3] OpenAI also says the model speaks with more natural and consistent accents across languages, and describes the API as improved in multilingual speech and voice fidelity.[1][4] Teams assessing it should test the languages, speech patterns and details relevant to their own requirements.
Delegating complex requests
OpenAI says GPT-Live-1.1 can delegate complex workflows to a backend model chosen by the developer.[2] The company says this delegation can happen without breaking the flow of conversation, and that requests needing more reasoning can be delegated to a developer-selected backend model.[2][4] OpenAI also says the model completes tasks more reliably.[1] The developer’s choice of backend is part of the described setup, so teams should assess how delegation fits their workflow and what model or tools it requires. Backend model and tool usage are billed separately from live-session rates.[3]
Endpoint and session rates
OpenAI identifies the model as `gpt-live-1.1` and lists support on the Live endpoint at `v1/live/sessions`.[3] The listed rate is $0.05 per minute for an active live session and $0.02 per minute for a standby live session.[3] Backend model and tool usage is billed separately.[3] Include the separate backend and tool charges when evaluating a proposed implementation.
Developers can now evaluate GPT-Live-1.1 for app-based voice experiences, with documented audio and text inputs and outputs. The separate active, standby, backend-model and tool charges make usage patterns and workflow design relevant to cost planning.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
19 November 2026
OpenAI makes GPT-Live-1.1 available in its API
Sources
- OpenAI Developers on X: \"GPT-Live-1.1 is now available in the API. It speaks with more natural and consistent accents across languages, supports lower idle costs with a new standby option, and completes tasks even more reliably.\" / Xx.com
- OpenAI Developers on X: \"Sound familiar? GPT-Live-1.1 has powered voice in ChatGPT Plus and Pro since October 15th. Now it can power your apps. Build better voice agents that can reliably understand precise details like letters and numbers, follow instructions, and delegate complex workflows to th… / Xx.com
- GPT-Live-1.1developers.openai.com
- GPT-Live-1.1 system card - OpenAI Deployment Safety Hubdeploymentsafety.openai.com
