Thanks for your reply. But it's to big for my PC. In PC around 1.5GB models got 20 token/s , which is too low for agentic workflow.