LFM2.5-Audio-1.5B
A tiny (1.5B) end-to-end speech and text model that does interleaved speech-to-speech chat and runs on CPU via GGUF. Ideal for on-device or edge voice, as long as your company is under the $10M revenue licence threshold.
Overview
Best for: On-device or edge voice assistants for small companies, offline kiosks, prototypes.
At a glance
LFM Open License v1.0: commercial use only for entities under $10M annual revenue. English (separate Japanese variant). Runs on CPU via GGUF.
Speech
Speech (interleaved with text)
English (Japanese variant available).
Not documented.
Designed for low-latency real-time conversation; no figure.
Self-hosted / on-device.
Self-hosted.
Runs on CPU via GGUF; a small GPU speeds it up (bfloat16, flash-attn optional).
LFM Open License v1.0: commercial use only for entities under $10M annual revenue; above that, use is not licensed.
Features
- interleaved speech-to-speech chat
- sequential ASR/TTS modes
- CPU inference via GGUF
Pricing
No public price list.
No licence fee; you pay for GPU time.
n/a (self-hosted)
Free tier: Open weights
Source: huggingface.co
Setup
- Check the licence threshold first.
- pip install liquid-audio and launch the demo, or use the GGUF build with llama.cpp on CPU.
Endpoint
Local Gradio demo on port 7860Authentication
n/a
Quick start python
pip install liquid-audio
pip install "liquid-audio[demo]"
liquid-audio-demo # serves on http://localhost:7860
# CPU: use LiquidAI/LFM2.5-Audio-1.5B-GGUF with llama.cppWritten from the current docs. Check the vendor's SDK version before you ship.
Warnings
Revenue cap in the licence
Commercial use is licensed only if you (or your legal entity) have under $10,000,000 annual revenue. Larger companies need a separate deal with Liquid AI.
Tiny model limits
1.5B parameters means weak knowledge and reasoning; pair with retrieval or keep tasks narrow.
English only
Main model is English; a separate JP model exists.
No full-duplex server
Interleaved generation is turn-based; barge-in handling is up to you.
Plus 3 warnings that apply to all open models APIs. See category warnings.
Limits
- Small model: limited knowledge/reasoning
- English only (main model)
Models and products
| Name | Status |
|---|---|
| LiquidAI/LFM2.5-Audio-1.5B (+ GGUF, ONNX) | GA |
| LiquidAI/LFM2.5-Audio-1.5B-JP | GA |
| LiquidAI/LFM2-Audio-1.5B | GA |
Docs and sources
Docs
Sources used
- huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
- huggingface.co/LiquidAI/LFM2.5-Audio-1.5B/blob/main/LICENSE
- huggingface.co/api/models?author=LiquidAI&search=Audio
Latency numbers; exact llama.cpp invocation.