Kyutai's full-duplex speech-to-speech foundation model in its male-voice variant, which models the user's and its own speech as parallel streams and reaches about 200 ms latency in practice.
Qwen/Qwen3-Omni-30B-A3B-Instruct
Indexed from Hugging FaceLinked to its home on Hugging Face. FrontierOS indexes it; nothing is hosted or mirrored here.ModelSpeech to speech35B paramsEst. DatacenterEstimated minimum compute for inference: Server-class GPU with 40 GB or more. Not a benchmark.Apache 2.0
About
A mixture-of-experts omni model with about 3B active parameters that takes text, image, audio, and video as input and replies in both text and streaming speech.
Why it matters for physical AI
A single open model that can see, hear, and talk back is a practical brain for a multimodal robot assistant on a workstation or server.
Summary written by FrontierOS in our own words. The upstream card and license govern use. Metadata checked 27 Sep 2026. Read the full card on Hugging Face
Files, weights and version history live on Hugging Face. FrontierOS indexes this repo; it does not host or mirror it.
hf download Qwen/Qwen3-Omni-30B-A3B-InstructUses the Hugging Face CLI (pip install -U huggingface_hub). This downloads the whole repo; add --include "<pattern>" to fetch only the files you need.
Full-duplex or cascaded voice for robots?
Full-duplex models such as Moshi can listen while they talk, which makes interruptions and backchannels feel natural. Cascaded stacks are easier to debug and swap piece by piece.