GPT Realtime 2.1 vs Cascade

Cascade is more reliable, by 3.7 points. GPT Realtime 2.1 passed 65 of 82 scenarios on all three runs, and Cascade passed 68. A scenario only counts when every run of it passed.

On reliability, no: the cascade passes more scenarios on every run. Cascade is ahead on reliability, success rate, data accuracy and calls where the agent went quiet, and GPT Realtime 2.1 on interruption and response time. The two tie on agent responses. Both ran as the same agent, with the same prompt and tools.

Cascade, by 2.1 points. GPT Realtime 2.1 saved everything exactly right on 92.0% of the calls that save data, and Cascade on 94.1%. On Booking calls, 95.9% against 98.8%. On Medicare calls, 81.8% against 81.8%. A call counts only when nothing is missing, wrong, extra or saved twice.

Yes. GPT Realtime 2.1 starts replying 0.23s sooner at the median. GPT Realtime 2.1 takes 1.94s at the median and 2.46s at p90; Cascade takes 2.17s and 2.51s. Both are measured on the call audio, from the caller finishing to the agent starting to reply.

GPT Realtime 2.1, by 0.02 on a 0 to 5 scale. GPT Realtime 2.1 scores 4.98 and Cascade 4.96 for how well the agent handled being talked over.

Cascade, by 2.4 points. The agent went quiet for 10 seconds or more on 3.3% of GPT Realtime 2.1 calls and 0.8% of Cascade calls.

There is no verified per-minute rate for Cascade, so cost cannot be compared yet.

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, 3,690 calls across the benchmark. A call passes only when the agent said the right things and saved the right data. The cascade runs the same agent for reference and is never ranked.