Smallest AI is ahead on infrastructure reliability, tool call accuracy, voice tone and clarity, interruption handling and response time, and Vapi on repeatable reliability and task completion.
Last updated Methodology by Luis Ojeda
Pick Smallest AI for calls that connect and stay up (13.41 points higher), accurate tool calls (0.56 higher), how the agent sounds (0.29 higher), handling interruptions (0.18 higher) and fast replies (0.37s faster). Pick Vapi for reliability across repeated runs (3.66 points higher) and completing the caller's task (3.25 points higher).
Bars share one scale across all 10 platforms. The place next to each figure is its rank in the field.
Share of the 82 scenarios that passed on all three runs.
Vapi, 3.66 points higher
Calls where the expected outcome was fully reached.
Vapi, 3.25 points higher
Calls with no connection, audio or platform failure.
Smallest AI, 13.41 points higher
Mean score for calling the right tool with the right arguments.
Smallest AI, 0.56 higher
Mean score for how clear and natural the agent sounds.
Smallest AI, 0.29 higher
Mean score for yielding and recovering when the caller cuts in.
Smallest AI, 0.18 higher
Mean time for the agent to start replying after the caller stops.
Smallest AI, 0.37s faster
246 calls per platform: 82 scenarios, 3 runs each.
Vapi is more reliable, by 3.66 points. Smallest AI passed 56.10% of the 82 scenarios on all 3 runs and Vapi passed 59.76%. A scenario only counts when every run of it passed.
Vapi is more reliable, by 3.66 points. Smallest AI passed 56.10% of the 82 scenarios on all 3 runs and Vapi passed 59.76%. A scenario only counts when every run of it passed.
Smallest AI starts replying sooner, 0.37s faster. Mean response time is 2.71s for Smallest AI and 3.08s for Vapi.
Smallest AI reached the expected outcome on 94.31% of calls and Vapi on 97.56%. Vapi is ahead on task completion.
Both ran the same Appointment and Medicare agents through the same 82 caller scenarios, 3 times each, with the same evaluators and mock tools. Only the platform and its speech and model components changed.
Cekura runs these simulated calls on your agent and scores every one.
