832
ELOELO Rating
12th of 12
11.4%
Pass³Reliability (Pass³)
12th of 12
2.76s
LatencyLatency P50
11th of 12
Higher wins more head-to-head calls (1000 = average); further left replies faster
Each model took the same recorded human callers on live calls. Pick a measure to see where Amazon Nova 2 Sonic ranks.
Rating from head-to-head calls: on every call each pair of models is compared, the better call wins. 1000 is the average model. Higher is better.
Pick a skill family. The shape is this model on every skill in it; the dashed line is the median of the other models.
Share of calls where the skill was needed and handled, strongest first. Hover a point on the chart for both values.
Banking and healthcare callers. Each shows what the caller needed, what made the call hard, and how the model handled it.
Finished 2 of 6 steps
Handled well: Emotion Recognition, Overlapping-Speech Recovery, Consent Verification
Struggled with: Multi-Speaker Handling, Tool-Latency Handling, Human Handoff
Listen to the callFinished the task
Handled well: Side-Speech Rejection, Tool-Latency Handling, Mid-Utterance Pause Tolerance
Listen to the callFinished 4 of 7 steps
Handled well: Degraded-Channel Robustness, Multi-Speaker Handling, Overlapping-Speech Recovery
Listen to the callFinished 2 of 7 steps
Handled well: Waiting Through Hold, Backchannel vs Barge-in Detection, Redundant Re-prompt Avoidance
Struggled with: Multi-Speaker Handling, Human Handoff, Consent Verification
Listen to the callFinished 1 of 3 steps
Handled well: Mid-Utterance Pause Tolerance, Hallucinated-Success Avoidance, Numeric Read-back
Struggled with: Tool-Latency Handling, Premature-Response Avoidance, Consent Verification
Listen to the callFinished 2 of 5 steps
Handled well: Cancellation Handling, Redundant Re-prompt Avoidance, TTS Text Normalization
Struggled with: Tool-Latency Handling, Consent Verification, Hallucinated-Success Avoidance
Listen to the callFinished 0 of 1 steps
Handled well: Endpointing, Premature-Response Avoidance, Redundant Re-prompt Avoidance
Listen to the callFinished 3 of 7 steps
Handled well: Backchannel vs Barge-in Detection, Redundant Re-prompt Avoidance, Hallucinated-Success Avoidance
Struggled with: Low-Confidence Confirmation, Consent Verification
Listen to the callAmazon Nova 2 Sonic starts replying in 2.76s on a typical turn and 3.62s on the slowest turn in twenty, 11th fastest of 12 at the median.
210 live calls: 70 scenarios, each run three times. The median call ran 6:28.
70 scenarios, the same for every model.
Amazon Nova 2 Sonic has an ELO rating of 832, 12th of 12 ranked models. It passed 8 of 70 callers on all three runs, a Pass³ of 11.4%.
Amazon Nova 2 Sonic has an ELO rating of 832, 12th of 12 ranked models. It passed 8 of 70 callers on all three runs, a Pass³ of 11.4%.
Amazon Nova 2 Sonic starts replying 2.76s after the caller finishes at P50, 3.40s at P90 and 3.62s at P95: 11th fastest of 12 at P50, measured on the call audio.
Amazon Nova 2 Sonic falls furthest behind on Tool Latency and Errors: 31%, against a median of 75%.
Amazon Nova 2 Sonic has a task completion rate of 13.8%, 10th of 12. A call counts only when every step the caller needed was done.
There is no verified per-minute rate for Amazon Nova 2 Sonic, so it is left out of the cost ranking.
Cekura ran Amazon Nova 2 Sonic as the whole agent with the same prompt, tools and phone connection as every other model. It took the same 70 recorded real human callers on live calls, three times each (210 calls), and each call was graded on 25 skills across Hearing, Turn Taking, Task Handling and Speaking.
Cekura runs these simulated calls on your agent and scores every one.
