Is Phonic-1 state of the art?Pending benchmarks.Announced Oct 29, 2026; no independent results yet.
Phonic's speech-to-speech model for phone and voice agents: it hears the caller's audio directly and answers in speech, without a separate transcription and text-to-speech step. Phonic says it can call tools while the caller is still talking and works in 50+ languages. Available now through Phonic's API at $0.07 per minute ($0.065 on the Pro plan), and through LiveKit Agents. No open weights.
The details so far
- Lab
- Phonic
- Announced
- Oct 29, 2026
- API
- Live · phonic-1
- Open weights
- Closed
- Type
- End-to-end speech-to-speech model for voice agents (audio in, audio out), with tool calls
- Price
- $0.07 per minute (Starter); $0.065 per minute on Pro with a $200 monthly minimum; no per-token or LLM pass-through charges
- Access
- Phonic API (WebSocket, model ID phonic-1), TypeScript and Python SDKs, CLI and a no-code agent builder; a Phonic plugin for LiveKit Agents
- Languages
- 50+, with switching mid-conversation in the same voice (per Phonic)
- Earlier Phonic models
- phonic-0.5 and merritt, still selectable in Phonic's API
- Artificial Analysis (Oct 29)
- Phonic 1 (High): Speech to Speech Index 87.1; Tau Voice 79.9%; Big Bench Audio 98.9%; time to first audio 0.61 s; $4.20 per hour of input audio
- Deployment
- Phonic Cloud, single-tenant, VPC peering or self-hosted (enterprise); SOC 2 Type II, HIPAA BAA available
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: the only speech-to-speech model that can call tools and act while the caller is still speaking.
Lab claim: responds in about 600 ms from speech in to speech out.
Lab claim: 83.2 on Sierra's multilingual full-duplex τ³-Voice, vs 79.9 for GPT-Live-1 and 68.5 for Gemini 3.8 Live.
Sources
LABPhonic blog: Introducing Phonic-1 and our $27.5M Series Aphonic.ai/blog/phonic-1 · Oct 29, 2026LABMoin Nadeem (Phonic) announcement on Xx.com/moinnadeem/status/2115830871014183088 · Oct 29, 2026LABPhonic post on Xx.com/phonic_co/status/2115838823183827310 · Oct 29, 2026LABPhonic pricingphonic.ai/pricingLABPhonic docs: introductiondocs.phonic.co/docs/get-started/overviewPRESSArtificial Analysis results for Phonic 1 (High) on Xx.com/ArtificialAnlys/status/2115836126925901915 · Oct 29, 2026Tracking continues until an independent board lists it.