Phonic-1
Phonic's speech-to-speech model for agent tasks: 50+ languages, tool calls while the caller speaks, and a claimed #1 on an AA index.
Overview
Phonic-1 is a speech-to-speech model from Phonic, announced October 29, 2026 for agentic voice tasks. It can call tools while the caller is still speaking, so answers are ready when they stop; Phonic cites a 44-minute hotel-booking call with 123 tool calls. It is trained on 50+ languages and can switch languages mid-conversation with the same voice, scoring 83.2 on Sierra's multilingual tau-3-Voice versus 79.9 for GPT-Live-1 and 68.5 for Gemini 3.8 Live. Phonic says it ranked first on the Artificial Analysis Speech to Speech Index, and its own blind tests with Podonos preferred it over GPT-Live-1 (61.5% to 32.0%) while human voice actors beat it (57.0% to 32.0%). All figures are company-reported.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Tool calls issued while the caller is still talking
- 50+ languages with mid-conversation language switching in one voice
- 83.2 on tau-3-Voice multilingual (company-reported; GPT-Live-1 79.9)
- Claimed first place on the Artificial Analysis Speech to Speech Index
- 1.1 s reply time in a travel-request demo (GPT-Live: 15.7 s in the same demo)
Use cases
- Voice agents that book travel, hotels or appointments through backend tools
- Multilingual customer-support calls
- Replacing cascaded speech-to-text, LLM and text-to-speech stacks
Pricing
No standard per-minute rate published. Sign-ups by November 30 get $250 in free credits; customers switching from cascaded stacks get three months free.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 29 Oct 2026
- Record updated
- 2 Nov 2026
- Record last checked
- 2 Nov 2026