Moshi Moshiko
Moshiko is one of Kyutai's released Moshi speech-dialogue models. It processes the user's audio while generating its own speech, so a conversation does not require alternating complete recordings.
Research release
Listening continues during a reply
The model predicts parallel audio streams and an internal text stream. That design supports spoken overlap and interruptions. It differs from a pipeline that first transcribes a complete turn, generates text, then reads it aloud.moshiko-pytorch-bf16: Kyutai model card
The release is a speech research model
The card notes limited performance on complex requests and no tool support. PyTorch, Candle and MLX packages, including quantized variants, provide different ways to run this voice; they do not establish separate trained models.moshiko-pytorch-bf16: Kyutai model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | kyutai/moshiko-pytorch-bf16kyutai/moshiko-mlx-q4kyutai/moshiko-candle-bf16kyutai/moshiko-candle-q8kyutai/moshiko-mlx-bf16kyutai/moshiko-mlx-q8kyutai/moshiko-pytorch-q8 |
| Language | English.moshiko-pytorch-bf16: Kyutai model card |
| Weight license | CC BY 4.0.moshiko-pytorch-bf16: Kyutai model card |
Related models
A streaming neural audio codec that compresses speech into tokens and reconstructs sound from them.
Launch date not confirmedResearch releaseDownloadable- Moshi Moshika
Kyutai · Moshi
A Moshi speech-dialogue model that listens while speaking, using Kyutai's Moshika voice.
Launch date not confirmedResearch releaseDownloadable