gpt-realtime-2.1
by OpenAI · Released 2026
OpenAI's latest realtime speech-to-speech model. Improves gpt-realtime-2 with better alphanumeric recognition, silence/noise handling, interruptions, and configurable reasoning effort.
gpt-realtime-2.1
Powered by OpenAI · Realtime multimodal
Context Window
128K
Parameters
Not disclosed
Max Output
N/A
Category
LLM Chat
Overview
`gpt-realtime-2.1` is OpenAI's latest realtime speech-to-speech model, updating gpt-realtime-2 with improved alphanumeric recognition, better silence and noise handling, smoother interruption behavior, and configurable reasoning effort for tool-driven voice agents. 128K text context. Voice-agent only (WebSocket).
Audio in/out rates match gpt-realtime-2 ($32/$64 per 1M); text output is $24/1M. On CallMissed it bills against active call minutes — roughly $0.375/min.
Use it when you want the newest realtime quality with the strongest recognition and interruption handling plus tunable reasoning.
Pricing
| Metric | Price |
|---|---|
| Input /1M tokens | ₹400.0000 |
| Output /1M tokens | ₹2400.0000 |
1 credit = ₹1 = $0.01 USD. Prices shown from provider; CallMissed passes through with ~35% markup.
Key Highlights
- Latest realtime
- Better recognition + interruptions
- Configurable reasoning
- $0.375/min
Technical Details
- Model id: gpt-realtime-2.1
- Voice-agent WebSocket only
- Configurable reasoning effort
- ~$0.375 per active call minute
Strengths
- Best-in-line recognition + interruption handling
- Configurable reasoning
- 128K context
Limitations
- Not available on chat completions
Use Cases
API Example
# Create a voice session with llm_model=gpt-realtime-2.1 via POST /v1/voice/sessions
Endpoint: WebSocket /v1/voice/sessions · Model ID: gpt-realtime-2.1
Try gpt-realtime-2.1 now
Get 1000 free API credits on signup. No credit card required.