Full Duplex Models: Moshi and the New Voice AI Paradigm
Source
Julia Turc
Author
Julia TurcTop Viber
Date
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
RAG — Retrieval-augmented generation — fetching relevant documents first and pasting them into the model's context so it answers from your data instead of memory.
Why it matters
If you're building voice agents, this lays out why full-duplex speech-to-speech models behave differently from ASR→LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition →→TTS stacks — latency, interruption handling, and the open problems with tool useA model's ability to call external functions — run code, search the web, edit files — instead of only generating text.Full definition → and RAGRetrieval-augmented generation — fetching relevant documents first and pasting them into the model's context so it answers from your data instead of memory.Full definition → — using Moshi as an open-source reference you can actually inspect and run.
Key quotes
“They're not turn-based. They're more like time-based interaction where they're continuously taking in audio, text, video, and continuously providing output.”
Mira Murati
“It's almost ideological, I would say, right? In a way, when you impose by your human hands a cascade in several gesture rings or signal network, you know, you go against the philosophy of machine learning.”
Nao
“Being a bit pessimistic, I would say 1 or 2 years.”
