---
title: LFM3-Audio-2B-MLX-4bit
canonical_url: "https://www.modelscope.cn/models/LiquidAI/LFM3-Audio-2B-MLX-4bit"
md_url: "https://www.modelscope.cn/models/LiquidAI/LFM3-Audio-2B-MLX-4bit.md"
repository: LiquidAI/LFM3-Audio-2B-MLX-4bit
last_updated: 2026-10-27
license: other
pipeline_tag: any-to-any
tasks:
  - any-to-any
model_type:
  - lfm3_audio
architectures:
  - Lfm3AudioForConditionalGeneration
base_model:
  - LiquidAI/LFM3-Audio-2B
base_model_relation: quantized
parameters: 621.3M
tensor_type:
  - U32
  - BF16
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
  - ar
  - da
  - de
  - el
  - fi
  - fr
  - he
  - it
  - ja
  - ko
  - nl
  - pl
  - pt
  - es
  - sv
  - tr
  - zh
downloads: 12
stars: 0
tags:
  - liquid
  - lfm3
  - mlx
  - audio
  - speech
---

# LFM3-Audio-2B-MLX-4bit

> LFM3-Audio-2B-MLX-4bit - LiquidAI 在 ModelScope 开源的模型。LFM3-Audio-2B-MLX-4bit

LiquidAI/LFM3-Audio-2B-MLX-4bit 是 ModelScope 魔搭社区上的 621.3M 参数any-to-any模型，采用 other 许可，基于 LiquidAI/LFM3-Audio-2B 构建。

- **Repository**: LiquidAI/LFM3-Audio-2B-MLX-4bit
- **License**: other
- **Tasks**: any-to-any
- **Parameters**: 621.3M
- **Base model**: LiquidAI/LFM3-Audio-2B
- **Tags**: liquid, lfm3, mlx, audio, speech
- **Downloads**: 12
- **Stars**: 0
- **Last updated**: 2026-10-27

Source: https://www.modelscope.cn/models/LiquidAI/LFM3-Audio-2B-MLX-4bit

---

<div align="center">
  <img
    src="https://huggingface.co/LiquidAI/LFM3-1.7B/resolve/main/assets/liquid-ai-banner.webp"
    alt="Liquid AI"
    style="width: 100%; max-width: 100%; height: auto; display: block; margin: 0;"
  />
  <div align="center" style="display: flex; justify-content: center; align-items: center; gap: 1.25em; margin: 8px 0 1.7rem; padding: 0; line-height: 1.3;">
    <a href="https://console.liquid.ai/dashboard" style="display: inline-flex; align-items: center; gap: 6px;"><img src="https://huggingface.co/LiquidAI/LFM3-1.7B/resolve/main/assets/icons/api.svg" width="16" height="16" alt="" aria-hidden="true" style="display: block; flex: 0 0 16px; width: 16px; height: 16px; margin: 0;" /><strong>API</strong></a>
    <a href="https://playground.liquid.ai/" style="display: inline-flex; align-items: center; gap: 6px;"><img src="https://huggingface.co/LiquidAI/LFM3-1.7B/resolve/main/assets/icons/try.svg" width="16" height="16" alt="" aria-hidden="true" style="display: block; flex: 0 0 16px; width: 16px; height: 16px; margin: 0;" /><strong>Try</strong></a>
    <a href="https://docs.liquid.ai/lfm/getting-started/welcome" style="display: inline-flex; align-items: center; gap: 6px;"><img src="https://huggingface.co/LiquidAI/LFM3-1.7B/resolve/main/assets/icons/docs.svg" width="16" height="16" alt="" aria-hidden="true" style="display: block; flex: 0 0 16px; width: 16px; height: 16px; margin: 0;" /><strong>Docs</strong></a>
    <a href="https://www.liquid.ai/blog/lfm3" style="display: inline-flex; align-items: center; gap: 6px;"><img src="https://huggingface.co/LiquidAI/LFM3-1.7B/resolve/main/assets/icons/blog.svg" width="16" height="16" alt="" aria-hidden="true" style="display: block; flex: 0 0 16px; width: 16px; height: 16px; margin: 0;" /><strong>Blog</strong></a>
  </div>
</div>

# LFM3-Audio-2B-MLX-4bit

4-bit MLX version of [LFM3-Audio-2B](https://huggingface.co/LiquidAI/LFM3-Audio-2B) for Apple Silicon. The weights take 1.58 GB, plus 0.15 GB for the audio codec in `audio_tokenizer/`.

The language backbone uses 4-bit affine quantization with groups of 64, except for the token embedding (also the output layer) and the value and feed-forward down projections in the layers that llama.cpp's Q4_K_M upgrades, which use 6 bits. The audio encoder, audio projector, depth decoder and codebook embedding stay in BF16. The audio codec is the original one. It is the smallest variant and loses some accuracy, so [8-bit](https://huggingface.co/LiquidAI/LFM3-Audio-2B-MLX-8bit) is the recommended trade-off.

Find architecture and usage details in the [original model card](https://huggingface.co/LiquidAI/LFM3-Audio-2B).

## Use with MLX-VLM
<hr style="margin: -0.7em 0 1.2em 0; border: 0; border-top: 1px solid rgba(127, 127, 127, 0.3);" />

Support for LFM3-Audio in [MLX-VLM](https://github.com/Blaizzy/mlx-vlm) is in progress. We will update this card when an MLX-VLM release supports it.

```bash
# Speech recognition
python -m mlx_vlm.generate --model LiquidAI/LFM3-Audio-2B-MLX-4bit \
  --system "Perform ASR in English." --prompt "" --audio <AUDIO PATH> \
  --temperature 0 --max-tokens 2048

# Text to speech
python -m mlx_vlm.generate --model LiquidAI/LFM3-Audio-2B-MLX-4bit \
  --system "Perform TTS in American English using a female voice." \
  --prompt "Hello there!" --output-modality audio --output tts.wav

# Speech-to-speech chat
python -m mlx_vlm.generate --model LiquidAI/LFM3-Audio-2B-MLX-4bit \
  --system "Respond with interleaved text and audio." --prompt "" \
  --audio <AUDIO PATH> --output-modality audio --max-tokens 1024 --output answer.wav
```

Speech output samples with `audio_temperature=1.0`, `audio_top_k=128` and `audio_repetition_penalty=1.1`, the defaults in `generation_config.json`. The original model card lists the voices, languages and other system prompts. More options in the [MLX guide](https://docs.liquid.ai/deployment/on-device/mlx).