- Reliability
- 58.5%
- Response time
- 2.31s
- Cost / min
- $0.0224
Gemini 3.8 Live Extended Thinking is ahead on success rate, data accuracy, calls where the agent went quiet, interruption and response time, and Deepslate Opal on reliability and agent responses.
Last updated Methodology by Dileep Chagam
For the wider field, see our pick of the best voice AI APIs for real-time audio.
Scenarios the agent passed on every one of its 3 runs.
Each model on its own: Deepslate Opal and Gemini 3.8 Live Extended Thinking
Deepslate Opal is more reliable, by 6.1 points. Deepslate Opal passed 48 of 82 scenarios on all three runs, and Gemini 3.8 Live Extended Thinking passed 43. A scenario only counts when every run of it passed.
Deepslate Opal is more reliable, by 6.1 points. Deepslate Opal passed 48 of 82 scenarios on all three runs, and Gemini 3.8 Live Extended Thinking passed 43. A scenario only counts when every run of it passed.
For a voice agent, Deepslate Opal, since it gets more scenarios right on every run. Gemini 3.8 Live Extended Thinking is ahead on success rate, data accuracy, calls where the agent went quiet, interruption and response time, and Deepslate Opal on reliability and agent responses.
Gemini 3.8 Live Extended Thinking, by 5.9 points. Deepslate Opal saved everything exactly right on 78.5% of the calls that save data, and Gemini 3.8 Live Extended Thinking on 84.4%. On Booking calls, 88.3% against 88.9%. On Medicare calls, 53.0% against 72.7%. A call counts only when nothing is missing, wrong, extra or saved twice.
No. Gemini 3.8 Live Extended Thinking starts replying 0.24s sooner at the median. Deepslate Opal takes 2.31s at the median and 2.88s at p90; Gemini 3.8 Live Extended Thinking takes 2.07s and 2.74s. Both are measured on the call audio, from the caller finishing to the agent starting to reply.
Gemini 3.8 Live Extended Thinking, by 0.01 on a 0 to 5 scale. Deepslate Opal scores 4.94 and Gemini 3.8 Live Extended Thinking 4.95 for how well the agent handled being talked over.
Gemini 3.8 Live Extended Thinking, by 0.8 points. The agent went quiet for 10 seconds or more on 6.1% of Deepslate Opal calls and 5.3% of Gemini 3.8 Live Extended Thinking calls.
There is no verified per-minute rate for Gemini 3.8 Live Extended Thinking, 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, 3,936 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.