Skip to content

Out of Memory

Go down one model size.

That solves more Gemma 4 setup failures than changing runtimes, changing prompts, or re-downloading the same model again.

  • load hangs forever
  • app crashes on load
  • model initializes but never becomes usable
  • the machine starts swapping heavily and everything becomes unusable
  1. Try a smaller model.
  2. Try a lighter quantized variant if your runtime supports it.
  3. Close other memory-heavy apps.
  4. Use a runtime that makes hardware usage more obvious.

Do not assume a runtime bug first. Most of the time the model simply does not fit cleanly on the chosen hardware.