970  Downloads Updated  3 days ago

LFM2.5-8B-A1B is a fast, memory-efficient hybrid MoE language model, designed for high-quality on-device AI and edge deployment. Unsloth Dynamic quants available!