OpenAI is ahead on reliability, success rate, agent responses, data accuracy, interruption and cost, and Google on calls where the agent went quiet and response time.
Last updated Methodology by Dileep Chagam
Each row sets Google's best model against OpenAI's best on that measure. For the wider field, see our pick of the best voice AI APIs for real-time audio.
OpenAI is ahead, 17.1 points higher. The best Google result is Gemini 3.1 Flash Live Preview at 62.2%, and the best OpenAI result is GPT Realtime 2.1 at 79.3%, on reliability.
OpenAI is ahead, 17.1 points higher. The best Google result is Gemini 3.1 Flash Live Preview at 62.2%, and the best OpenAI result is GPT Realtime 2.1 at 79.3%, on reliability.
Google is faster, 0.03s sooner. The best Google result is Gemini 3.8 Live at 1.91s, and the best OpenAI result is GPT Realtime 2.1 at 1.94s, on response time.
Google: Gemini 3.8 Live Extended Thinking, Gemini 3.8 Live, Gemini 3.1 Flash Live Preview. OpenAI: GPT Realtime 2.1, GPT-Live 1 + GPT-6 Sol, GPT Realtime 2.1 Mini. Every model ran the same tests on Speech-to-speech.
Cekura runs these simulated calls on your agent and scores every one.