← Back to Voice models
OpenAINotable
GPT-Realtime-2
Released May 7, 2026
OpenAI's first Realtime API voice model with configurable reasoning, for speech-to-speech agents. It is still listed alongside its successor, GPT-Realtime-2.1.
- Price per minute
- —
- Pricing
- Per 1M tokens: text $4 in / $0.40 cached / $24 out; audio $32 in / $0.40 cached / $64 out; image $5 in / $0.50 cached.
- Inputs
- audio, text, image
- Tool calling
- Yes
- Languages
- —
- Voices
- 10
- Big Bench Audio
- 96.6%
- Access
- Generally available
- Licence
- Proprietary
Strengths & weaknesses
Strengths
- Highest Big Bench Audio of any OpenAI model on AA, 96.6% at High
- Configurable reasoning effort and function calling
- 128K context, 32K max output
Weaknesses
- Superseded by GPT-Realtime-2.1 at the same price
- Big Bench Audio falls to 71.8% at minimal effort (AA)
Notable: frontier-lab clause. Not deprecated on OpenAI's deprecations page, but GPT-Realtime-2.1 updates it at identical pricing; consider dropping it if the directory keeps only the newest generation. AA Big Bench Audio: High 96.6%, Medium 93.3%, Minimal 71.8%; S2S Index 73.6 (High).
Links
API model strings
- openai
gpt-realtime-2