LaunchAI ModelsOctober 22, 2026

LiquidAI releases LFM3-Omni-2B omni model

Read original source →huggingface.co

LiquidAI's LFM3-Omni-2B accepts visual, audio, or text and emits interleaved text and audio tokens end-to-end, with no separate ASR or TTS models. Custom WebGPU kernels run it at ~280 tok/s on an M4 Max in-browser; a GGUF build landed with same-day llama.cpp support.

How this story unfolded

3 days · 1 report · 2 community posts · from Oct 27

  1. Oct 27
  2. Oct 28
  3. Oct 30

More stories today

Open the live feed