As reported by marktechpost, Liquid AI has formally introduced the LFM2.5 model family. Built upon the LFM2 framework, which has been fine-tuned specifically for edge computing devices, this collection encompasses both Base and Instruct iterations, each boasting 1.2 billion parameters. Moreover, it expands to encompass multimodal variations, including models tailored for Japanese language processing, as well as vision-language and audio-language fusion models. All these models are made accessible with their weights openly available on the Hugging Face platform, and they come with deployment assistance facilitated through the LEAP platform.
