Phonic Launches Phonic-1 Voice AI That Starts Working Before the Speaker Stops
Phonic announced the voice AI model Phonic-1 on October 29, 2026. The model processes speech directly and begins searching for information and calling external tools while the other person is still talking.
1 source 2h ago
Reporting from 1 source: GIGAZINE.
Why it matters
Phonic-1's pitch is latency, and the demo it published puts its answer time at roughly a fourteenth of GPT-Live's, which is the number that will decide whether voice agents get used for reservations and support calls rather than just chat.
Key Facts
- Phonic announced the voice AI model Phonic-1 on October 29, 2026.
- In a demo on the official site, Phonic-1 called five external tools during the utterance and answered 1.1 seconds after the utterance ended.
- OpenAI's GPT-Live called external tools after the utterance ended and took 15.7 seconds to answer in the same comparison.
- In a Podonos test, Phonic-1 responses were preferred over GPT-Live-1's in 61.5% of cases, while human voices were chosen 57.0% of the time against Phonic-1.
- Phonic-1 has been trained on more than 50 languages and can keep the same voice when the language is switched mid-conversation.
In a demo on the official site, Phonic-1 answered 1.1 seconds after the utterance ended, against 15.7 seconds for OpenAI's GPT-Live. Phonic says Phonic-1 is already available and trained on more than 50 languages.
Phonic-1 skips the usual voice-to-text-to-voice pipeline. It handles speech directly and starts the lookup or the external service call mid-sentence, so the pause after the speaker stops is shorter. The company's own demo has it calling five external tools during the utterance and answering 1.1 seconds after the speaker finished; GPT-Live, which calls tools after the utterance ends, took 15.7 seconds in the same comparison.
Naturalness is the weaker side. In a third-party test run by Podonos, Phonic-1 responses were preferred over GPT-Live-1's in 61.5% of cases, but against human voices the humans won 57.0% of the time, and Phonic says conversational naturalness still has room to improve.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google