Find where downloaded Gemma 4 weights are stored
getting started
// what it does
Ollama stores pulled models under ~/.ollama/models (blobs plus a manifest). Transformers and huggingface-cli cache to ~/.cache/huggingface/hub, and llama.cpp's -hf downloads land in ~/.cache/llama.cpp. Set HF_HOME or OLLAMA_MODELS to relocate them off a small root disk.
// shell
$ du -sh ~/.ollama/models$ du -sh ~/.cache/huggingface/hub// gotcha
A single 31B checkpoint is 20+ GB and the family has five sizes plus quantized variants, so pulling several tags quietly fills a disk. Audit with ollama list and du -sh, then reclaim space with ollama rm <tag>.