Phonic introduced Phonic-1, a speech-to-speech model for agentic tasks. It tops the Artificial Analysis Speech to Speech Index. The model achieves the highest task success rate at the lowest latency. Phonic-1 can act while the caller is still talking. It supports 50+ languages and switches mid-conversation. It scored 83.2 on Sierra’s multilingual full-duplex benchmark. A demo showed 123 tool calls in a 44-minute call. Phonic also announced a $27.5M Series A led by First Round. Total funding is now $31.5M. Phonic-1 is available today with $250 in free credits for new sign-ups by November 30.
What changed
Voice agents used cascaded pipelines with separate transcription, reasoning, and speech models.
What it unlocks
Acting on caller requests while they are still speaking.
- $27.5M Series A
- $31.5M total funding
- 50+ languages
- 83.2 multilingual full-duplex score
What you need to act on it
- sign up by November 30 for free credits
Sources