ZMIME

GPT-Live-1.1 vs MAI Realtime 1

Compare documented capabilities and access. A dash means a comparable figure is unavailable.

Model specifications
SpecificationGPT-Live-1.1OpenAIMAI Realtime 1Microsoft
TaskReal-time voice conversationReal-time voice conversation
Published capabilitiesImproves multilingual pronunciation, accent consistency and transcription over GPT-Live-1. Supports backend delegation during a conversation.Handles interruptions and background speech and can delegate difficult requests to a reasoning engine.
AccessChatGPT Voice from October 15. Live API access followed on November 19.MAI Labs demo and limited-access Microsoft Foundry preview
ResultStreaming speech and textSpoken replies
Billing terms$0.05 per active minute and $0.02 per paused WebRTC standby minute. Backend model and tool usage is billed separately.$0.01 per audio minute for base speech-to-speech. Additional reasoning costs $1 per million input tokens and $5 per million output tokens.
SourcesPublisher documentation · Additional documentationPublisher documentation
InputsAudio, TextAudio
WeightsWeights unconfirmedWeights unconfirmed
ReleasedOct 15, 2026Nov 17, 2026