LFM

What it does

LFM (Liquid Foundation Models) is Liquid AI's model family built to run on the device rather than in the cloud. The LFM2.5 line spans 230M to 24B parameters (the 24B-A2B mixture-of-experts activates about 2B per token), with vision, audio and encoder models alongside, and ships in GGUF, MLX and other formats for llama.cpp, MLX, ONNX and vLLM. Weights are on Hugging Face under Liquid AI's own license; an API and fine-tuning are offered too.

Compare with

• vsChatGPT
• vsStep
• vsGemma

People who use it also use

Not enough marks yet to tell. When people who use LFM mark the rest of their stack, it shows up here.