OpenAI: gpt-realtime-mini API
A cost-efficient realtime multimodal model released by OpenAI in October 2025. As the lighter version of the gpt-realtime flagship, it is optimized for low-latency speech-to-speech interactions. It processes audio streams natively—capturing tone, inflection, and emotion—with a human-like response time of ~300ms.
- Context window: 32,000 tokens
- Max output: 4,096 tokens
- Input: audio, text
- Output: text, audio
- Tool calling: Supported
- Released: 2025-12-14
- Knowledge cutoff: 2024-09-30
Frequently Asked Questions
What is the context window of gpt-realtime-mini?
gpt-realtime-mini supports a context window of up to 32,000 tokens.
Does gpt-realtime-mini support function calling?
Yes. gpt-realtime-mini supports tool / function calling.
What is the knowledge cutoff of gpt-realtime-mini?
The knowledge cutoff of gpt-realtime-mini is 2024-09-30.