Skip to main content
Audio modelActive

GPT-Realtime 2.1 Mini

OpenAI

Released
July 6, 2026
Data date
October 25, 2026
Model class
Real-time audio model
Access
API

GPT-Realtime 2.1 Mini is OpenAI’s real-time model for audio and voice conversations. It accepts audio, text, and images, returns audio and text, and has a 128,000-token context with up to 32,000 output tokens.

OpenAI released it on July 6, 2026. Per 1 million input tokens, text costs $0.60, cached text $0.06, audio $10, cached audio $0.30, images $0.80, and cached images $0.08; per 1 million output tokens, text costs $2.40 and audio $20.

Specifications and access

SpecificationValue and source
Model class
Real-time audio modelSource
Access
APISource
API model ID
gpt-realtime-2.1-miniSource
Input
Audio, image and textSource
Output
Audio and textSource
Task
Voice dialogueSource
Supported details
Real-time voice dialogue with tool useSource
Context window
128,000 tokensSource
Maximum output
32,000 tokensSource
Knowledge cutoff
2024-09-30Source