# New audio model: LFM3-Audio-2B — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/a94cb023b032b28a

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-27 21:14 ingested 2026-10-28 02:38

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xh50at/new_audio_model_lfm3audio2b/)

## retained text



```
One more great model from Liquid AI!

Good for edge:

LFM3-Audio-2B is a **compact, fully featured audio model that runs locally on laptops and edge devices**. It takes text or audio as input and responds with text, speech, or both.

* **Voice chat:** ask questions aloud and receive streaming replies in text and speech. For text-only responses, the model can reason before answering.
* **Multilingual speech:** transcribe, synthesize, and translate speech across 18 languages.
* **Audio understanding and reasoning:** interpret recordings and answer questions about the speech, sounds, and events they contain.
* **Tool calling:** turn spoken requests into tool calls to control apps and devices, with text or speech alongside the calls.

Capabilities

* Speech recognition (ASR)
* Real-time speech recognition (LiveASR)
* Text-to-speech synthesis (TTS)
* Voice chat with spoken and text responses
* Speech-to-text chat with optional reasoning
* Audio understanding
* Speech translation
* Voice-activated tool calling
```



document fetched 2026-10-28 00:42 · canonical https://reddit.com/r/LocalLLaMA/comments/1xh50at/new_audio_model_lfm3audio2b
