My conclusion after a few days of local llming is that qwen3.5 is the best model for nerds, gemma4 not very impressive and pretty bad at tool use out of the box, maybe this will improve
Running gemma 4 on ollama, latest llama.cpp or something else?
3
