Kyutai (Moshi / Unmute)
French nonprofit open-sourcing Moshi and Unmute real-time voice AI
Overview
Kyutai is a Paris-based nonprofit AI lab, backed by over $300 million from Xavier Niel, Rodolphe Saadé and Eric Schmidt, that open-sources full-duplex voice models — Moshi, the Unmute pipeline, and lightweight Kyutai TTS/STT — for developers to self-host.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Kyutai (Moshi / Unmute)
- Moshi: full-duplex, speech-native dialogue model, open weights
- Unmute: open pipeline adding real-time voice to any text LLM
- Kyutai TTS and STT models, open-source
- Kyutai Pocket TTS: 100M-param, CPU-capable, multilingual
- Mimi neural audio codec
- All code and weights self-hostable, no vendor lock-in
ASEAN Perspective
Kyutai (Moshi / Unmute) in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Kyutai is the rare well-funded lab that actually ships open weights: Moshi, released in July 2024, was one of the first speech-native, full-duplex dialogue models, and Unmute lets you bolt real-time listening and speaking onto any text LLM using fully open components. The January 2026 Kyutai Pocket TTS — 100 million parameters, runs on CPU, expanded to five additional languages by April — shows the team iterating on practical, deployable models rather than just research demos.
Hacker News reception at Moshi's launch was genuinely mixed: praised as a technical achievement and "refreshing" to see a European open-science lab compete, but reviewers also found it well behind GPT-4o on raw conversational intelligence, occasionally unstable in demos, and thin on tooling (no llama.cpp or Ollama support at launch). It's infrastructure for engineering teams willing to self-host, not a consumer product — a serious, low-cost building block rather than a ChatGPT rival.
What people say
Kyutai launched in November 2023 in Paris as a nonprofit research lab, backed by a €300 million (roughly $330 million) five-year commitment split between Iliad founder Xavier Niel, CMA CGM's Rodolphe Saadé, and former Google CEO Eric Schmidt through Schmidt Futures. It was billed at the time as Europe's first independent open-science AI lab, staffed early on with researchers who had come from Microsoft, Goldman Sachs and Meta.
Moshi, released in July 2024, was the lab's first major output: a speech-native dialogue model that processes audio directly instead of running the usual transcribe-then-generate-then-speak pipeline, which gives it low latency and some ability to pick up on tone. The full model and code are open-sourced on GitHub under kyutai-labs for local or self-hosted use.
Unmute, a separate open pipeline, adds real-time speech-to-text and text-to-speech around any existing text LLM rather than requiring a speech-native model, and every component in it is open-source. Kyutai TTS and STT ship as standalone models. The newest release, Kyutai Pocket TTS in January 2026, is a 100-million-parameter model light enough to run in real time on a CPU, and by April 2026 the team had extended it to five additional languages beyond English. Between June 2025 and February 2026 Kyutai also ran a public Voice Donation Project, collecting volunteer voice samples to expand its open TTS training data.
Reception on Hacker News at Moshi's launch split along predictable lines: enthusiasm that a European lab was "knocking it out of the park" by going fully open, against concrete complaints that the model lagged well behind GPT-4o on conversational intelligence, had rough moments in live demos, and shipped with no backend tooling support (no llama.cpp or Ollama integration at release). More than a year later, the follow-up releases — Pocket TTS, added languages, the voice-donation dataset — read as the team addressing exactly those practicality gaps rather than resting on the initial demo's buzz.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Kyutai (Moshi / Unmute)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Kyutai (Moshi / Unmute) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Kyutai (Moshi / Unmute) directly →
Spotted something out of date? Suggest an update →
Kyutai (Moshi / Unmute) in the news
Alternatives to Kyutai (Moshi / Unmute)
More in Video & Audio
