maternion/ lfm2.5:8b-Q4_K_XL

971 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!

tools thinking 8b
... /
params
3b15415e7162 · 53B
{
"repeat_penalty": 1.05,
"temperature": 0.2,
"top_p": 80
}