OpenAI Realtime API vs Gemini Live API
Pricing and features, side by side. Checked October 14, 2026.
| OpenAI Realtime API | Gemini Live API | |
|---|---|---|
| Starting price | gpt-realtime-2.1-mini: audio $10 in / $20 out per 1M tokens | Free tier; paid audio ~$0.005/min in, ~$0.018/min out (Gemini 3.8 Live) |
| Billing | Per token (audio, text, image); some per-minute products | Per token (per-minute equivalents published) |
| Best for | Developers building speech-to-speech voice agents directly on OpenAI models | Developers who want a low-latency speech-to-speech model from Google |
| Key features | Model works directly with audio, keeps conversation state and calls tools | Audio-to-audio models built for real-time voice agents |
| WebRTC in the browser or WebSocket on the server | Optional extended thinking | |
| Barge-in, turn detection and tool use in one session |
Both are speech-to-speech model APIs billed per token. OpenAI's gpt-realtime-2.1 costs $32 per 1M audio input tokens and $64 per 1M audio output tokens ($10 / $20 on gpt-realtime-2.1-mini); its GPT-Live 1 sessions are $0.05/min. Google's Gemini 3.8 Live has a free tier, and paid audio is $3.00 in / $12.00 out per 1M tokens, which Google lists as about $0.005/min in and $0.018/min out.
Pick OpenAI Realtime if you want OpenAI's realtime models with WebRTC or WebSocket sessions and built-in tool use.
Pick Gemini Live if you want a free tier and Google's published per-minute audio rates.
Sources: OpenAI Realtime API · Gemini Live API