[limondas]

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.

[denn-gubsky]

try latest gemma4:12b. It fits into 16Gb with 256K context window