Guava's Daytona tops four-agent voice index

Guava’s Daytona scored 58.88/100, first among four systems in the company’s Voice Index, just 1.77 points ahead of ElevenLabs paired with GPT-5.4; evaluators still preferred human agents in 91% of response-timing comparisons and 93% of interruption-recovery comparisons. Guava publishes and competes in the index, and no tested system reached 60, so the result is directional rather than human parity; its launch release says Daytona is available to Guava customers. Teams piloting voice automation should own interruption and end-to-end resolution tests before replacing human queues; our voice-agent budget analysis shows why token price is only one input.