Voice-agent developers get better tool calls, but audio tokens at $64 per million output are pricey, and a 2.1 update already exists.
Kenning's take
What it does
GPT-Realtime-2 is an OpenAI voice model for developers building live speech-to-speech agents, with adjustable reasoning and more reliable tool calls. It is used through the Realtime API. OpenAI released GPT-Realtime-2.1 as an update in July 2026.
- Speech-to-speech conversations
- Configurable reasoning effort
- Tool calling in live calls
- Text, audio and image input
- 128K-token context window
Who it suits. Developers building phone or app voice agents.
Key facts
- Category
- AI models
- Made by
- OpenAI
- Based in
- United States
- Released
- 7 May 2026
- How to get it
- OpenAI Realtime API (model gpt-realtime-2)
- Works on
- API
- In the UK
- Yes
- API price
- Text $4 per million input tokens and $24 output; audio $32 per million input tokens and $64 output
FAQ
When was GPT-Realtime-2 released?
OpenAI released it in the API on 7 May 2026.
Is there a newer version?
Yes, GPT-Realtime-2.1 was released on 6 July 2026.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.