- Reliability
- 70.7%
- Response time
- 1.66s
- Cost / min
- $0.0798
GPT Realtime 2.1 is ahead on reliability, success rate, agent responses, data accuracy, calls where the agent went quiet and interruption, and Azure Realtime on response time and cost.
Last updated Methodology by Dileep Chagam
Our guide covers testing OpenAI Realtime API voice agents.
Scenarios the agent passed on every one of its 3 runs.
GPT Realtime 2.1 is more reliable, by 8.5 points. Azure Realtime passed 58 of 82 scenarios on all three runs, and GPT Realtime 2.1 passed 65. A scenario only counts when every run of it passed.
GPT Realtime 2.1 is more reliable, by 8.5 points. Azure Realtime passed 58 of 82 scenarios on all three runs, and GPT Realtime 2.1 passed 65. A scenario only counts when every run of it passed.
For a voice agent, GPT Realtime 2.1, since it gets more scenarios right on every run. GPT Realtime 2.1 is ahead on reliability, success rate, agent responses, data accuracy, calls where the agent went quiet and interruption, and Azure Realtime on response time and cost.
GPT Realtime 2.1, by 5.9 points. Azure Realtime saved everything exactly right on 86.1% of the calls that save data, and GPT Realtime 2.1 on 92.0%. On Booking calls, 94.2% against 95.9%. On Medicare calls, 65.2% against 81.8%. A call counts only when nothing is missing, wrong, extra or saved twice.
Yes. Azure Realtime starts replying 0.28s sooner at the median. Azure Realtime takes 1.66s at the median and 2.22s at p90; GPT Realtime 2.1 takes 1.94s and 2.46s. Both are measured on the call audio, from the caller finishing to the agent starting to reply.
GPT Realtime 2.1, by 0.01 on a 0 to 5 scale. Azure Realtime scores 4.96 and GPT Realtime 2.1 4.98 for how well the agent handled being talked over.
GPT Realtime 2.1, by 0.8 points. The agent went quiet for 10 seconds or more on 4.1% of Azure Realtime calls and 3.3% of GPT Realtime 2.1 calls.
Azure Realtime is $0.0055 cheaper a minute. Azure Realtime costs $0.0798 a minute of call and GPT Realtime 2.1 $0.0853, at each vendor's published rates.
Cekura ran both as the whole agent in the same open-source Pipecat pipeline, with the same prompt, tools and connection. A simulated caller worked through all 82 scenarios on live calls, three times each, 3,936 calls across the benchmark. A call passes only when the agent said the right things and saved the right data.
Cekura runs these simulated calls on your agent and scores every one.