Phonic-1 answers 1.1 seconds after speech ends

Phonic-1 arrives alongside a $27.5M Series A raise, giving the company runway to push its speech-to-speech approach into commercial deployments. The core technical departure is that Phonic-1 does not wait for a finished sentence: it listens, starts external searches and tool calls mid-utterance, and only then speaks. That design also preserves vocal tone and pauses that are typically lost when audio is converted to text first.

Phonic's own comparison makes the gap concrete. In the published demo, Phonic-1 invoked five external tools while the user was still speaking and replied 1.1 seconds after the sentence ended, while OpenAI's GPT-Live invoked tools only after the sentence and took 15.7 seconds. Naturalness is a separate matter: in Podonos' third-party listening test, Phonic-1's responses were preferred 61.5% of the time over GPT-Live-1's, but human voices were still chosen 57.0% of the time over Phonic-1, and Phonic acknowledges room to improve. With availability already in place and target sectors named, the next question is how the latency advantage holds up outside a demo.