- Reliability
- 79.3%
- Response time
- 1.94s
- Cost / min
- $0.0853
GPT Realtime 2.1 is ahead on reliability, success rate, agent responses, data accuracy and interruption, and PolyAI Dialog-RSN-1 on calls where the agent went quiet, response time and cost.
Last updated Methodology by Dileep Chagam
Scenarios the agent passed on every one of its 3 runs.
GPT Realtime 2.1 is more reliable, by 14.6 points. GPT Realtime 2.1 passed 65 of 82 scenarios on all three runs, and PolyAI Dialog-RSN-1 passed 53. A scenario only counts when every run of it passed.
GPT Realtime 2.1 is more reliable, by 14.6 points. GPT Realtime 2.1 passed 65 of 82 scenarios on all three runs, and PolyAI Dialog-RSN-1 passed 53. 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 and interruption, and PolyAI Dialog-RSN-1 on calls where the agent went quiet, response time and cost.
GPT Realtime 2.1, by 19.0 points. GPT Realtime 2.1 saved everything exactly right on 92.0% of the calls that save data, and PolyAI Dialog-RSN-1 on 73.0%. On Booking calls, 95.9% against 97.7%. On Medicare calls, 81.8% against 9.1%. A call counts only when nothing is missing, wrong, extra or saved twice.
No. PolyAI Dialog-RSN-1 starts replying 0.57s sooner at the median. GPT Realtime 2.1 takes 1.94s at the median and 2.46s at p90; PolyAI Dialog-RSN-1 takes 1.36s and 1.87s. Both are measured on the call audio, from the caller finishing to the agent starting to reply.
GPT Realtime 2.1, by 0.13 on a 0 to 5 scale. GPT Realtime 2.1 scores 4.98 and PolyAI Dialog-RSN-1 4.85 for how well the agent handled being talked over.
PolyAI Dialog-RSN-1, by 3.3 points. The agent went quiet for 10 seconds or more on 3.3% of GPT Realtime 2.1 calls and 0.0% of PolyAI Dialog-RSN-1 calls.
PolyAI Dialog-RSN-1 is $0.0447 cheaper a minute. GPT Realtime 2.1 costs $0.0853 a minute of call and PolyAI Dialog-RSN-1 $0.0406, 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.