GPT-Live 1 + GPT-6 Sol vs GPT Realtime 2.1

GPT Realtime 2.1 is more reliable, by 3.7 points. GPT-Live 1 + GPT-6 Sol passed 62 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, data accuracy, calls where the agent went quiet, interruption and response time, and GPT-Live 1 + GPT-6 Sol on agent responses and cost.

GPT Realtime 2.1, by 3.0 points. GPT-Live 1 + GPT-6 Sol saved everything exactly right on 89.0% of the calls that save data, and GPT Realtime 2.1 on 92.0%. On Booking calls, 100.0% against 95.9%. On Medicare calls, 60.6% against 81.8%. A call counts only when nothing is missing, wrong, extra or saved twice.

No. GPT Realtime 2.1 starts replying 0.08s sooner at the median. GPT-Live 1 + GPT-6 Sol takes 2.02s at the median and 2.42s 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 0.09 on a 0 to 5 scale. GPT-Live 1 + GPT-6 Sol scores 4.89 and GPT Realtime 2.1 4.98 for how well the agent handled being talked over.

GPT Realtime 2.1, by 8.9 points. The agent went quiet for 10 seconds or more on 12.2% of GPT-Live 1 + GPT-6 Sol calls and 3.3% of GPT Realtime 2.1 calls.

GPT-Live 1 + GPT-6 Sol is $0.0300 cheaper a minute. GPT-Live 1 + GPT-6 Sol costs $0.0553 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, 3,690 calls across the benchmark. A call passes only when the agent said the right things and saved the right data.