Positioning and Release Date
GPT-4o Realtime is a preview model for real-time voice and text interaction. On June 3, 2025, OpenAI announced the snapshot gpt-4o-realtime-preview-2025-06-03 and made this version available in the Realtime API. (community.openai.com)
Capabilities and Uses
- Supports text and audio input and output, with real-time interaction via WebRTC or WebSocket, making it suitable for building voice assistants and real-time conversational agents. (developers.openai.com)
- Supports function calling. OpenAI states that this update improved instruction following, tool-calling consistency, and behavior when interrupted, and introduced a voice playback speed parameter. (developers.openai.com)
- Realtime API sessions can use the Traces dashboard to view processes such as audio input and output, tool calls, and interruptions. (community.openai.com)
Limitations and Status
- The context window is 32,000 tokens, the maximum output is 4,096 tokens, and the knowledge cutoff date is October 1, 2023. (developers.openai.com)
- The official model page lists image, video, structured output, and fine-tuning as unsupported. (developers.openai.com)
- This snapshot was shut down on May 7, 2026; the replacement model recommended by OpenAI in its deprecation documentation is
gpt-realtime-1.5. (developers.openai.com)
Pricing
As checked on November 18, 2026, the OpenAI model page still displayed the following prices in US dollars (USD), all per million tokens. Because the model has been shut down, these figures do not indicate that it can currently be accessed or purchased. (developers.openai.com)
| Modality | Input | Cached input | Output |
|---|---|---|---|
| Text | 5.00 USD | 2.50 USD | 20.00 USD |
| Audio | 40.00 USD | 2.50 USD | 80.00 USD |