GPT-Audio-1.5

GPT-Audio-1.5, released on 23 February 2026, is OpenAI's audio model for the Chat Completions API: it takes speech or text in and returns speech or text out, without the persistent WebRTC or WebSocket session that the Realtime models require. That single design decision is the point of the model. A developer who already sends JSON to a chat endpoint can add a voice reply by changing the request body, not the transport layer, which makes it the cheapest way into voice for existing applications — at the cost of the low latency and interruption handling that only a live session can give. The model holds a 128,000-token context, returns up to 16,384 tokens, streams its answer and calls functions. Audio costs 32 dollars per million input tokens and 64 per million output, while text stays at the usual 2.50 and 10 dollars. Its knowledge ends on 30 September 2024.

▸Videos

No videos yet.

▸Reviews