LiquidAI/LFM3-Audio-2B
AI model · https://huggingface.co/LiquidAI/LFM3-Audio-2B
First seen by Oriel on .
AI summary
LFM3-Audio-2B is a 2B-parameter local multimodal audio model supporting speech recognition, synthesis, translation, understanding, voice chat, and tool calling.
- It accepts text or 16 kHz audio and produces text, 24 kHz speech, or interleaved text and speech within a 32K-token context.
- An in-house audio encoder and LiquidCodec audio tokenizer improve speech quality over LFM2.5-Audio while enabling causal streaming and incremental decoding.
- The model supports speech transcription, synthesis, and translation across 18 listed languages and locales, with LiveASR currently limited to English.
- Inference variants target vLLM, SGLang, Transformers, llama.cpp, MLX-VLM, and ONNX Runtime across GPUs, CPUs, and Apple Silicon.
- It is designed for local agents, structured outputs, tool calling, and fine-tuning, but is not recommended for agentic coding or complex multi-step reasoning.
#audio models #speech recognition #text to speech #tool calling
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 228, latest observation ; +122 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 8,018, latest observation ; +7,316 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | LiquidAI/LFM3-Audio-2B | https://huggingface.co/LiquidAI/LFM3-Audio-2B Read discussion |