Ollama 0.40.13 brings Liquid LFM3 audio models to Windows and Linux
Ollama released v0.40.13 with Linux and Windows support for Liquid AI's LFM3 audio models, lfm3-audio and lfm3-omni. The ~2B-parameter models fit in 1.5–2.4 GB and handle speech recognition, translation and text-to-speech across roughly 18 languages locally.
- LFM3 audio models take 1.5–2.4 GB depending on quantization
- 16 kHz audio input, 24 kHz speech output, 32K-token context
- Served via OpenAI-compatible /v1/audio/transcriptions and /v1/audio/speech endpoints
- Text-only LFM3 checkpoint is about 1.1 GB at Q4_K_M
Read next
AI