Run these first
Common issues and fixes
'ollama' is not recognized as an internal or external command
This means ollama.exe is not on your PATH, or the terminal was not restarted after install. Fix:
If it still fails: Settings → Apps → confirm Ollama is installed. If not, download and install. For full PATH fix steps see Not recognized guide.
Port 11434 is already in use
Another process is using the port Ollama needs. Find and stop it:
See Port 11434 guide for more scenarios.
Model download is stuck or very slow
Ollama resumes interrupted downloads. If it appears stuck, check disk space and internet. To cancel and retry:
Also check: antivirus may be scanning large model files and slowing the download. Add the .ollama\models folder to your antivirus exclusion list.
GPU not being used (gpu layers: 0)
Ollama is falling back to CPU. Most common causes on Windows: outdated NVIDIA driver (needs 527+), GPU VRAM too low for the model, or DirectML not detected for AMD. See the full GPU Acceleration guide.
Out of memory error
The model does not fit in available RAM or VRAM. Solutions: (1) Use a smaller model or lower quantization — try ollama pull llama3:8b-instruct-q4_K_M instead of the default. (2) Close other GPU-heavy applications. (3) On CPU: ensure you have at least 8 GB RAM free.
Firewall or antivirus blocks Ollama
Ollama needs to bind to localhost:11434. When prompted by Windows Firewall, click Allow access for private networks. If a corporate policy blocks it, add an exception for ollama.exe in your antivirus/firewall settings.
ollama serve: listen tcp 0.0.0.0:11434: bind: address already in use
Ollama is already running in the background (check the system tray). You cannot run two instances at once. Right-click the tray icon → Quit, then reopen Ollama, or just use the existing running instance.
Error: model not found
The model name is wrong or not downloaded yet. Check available models: ollama list. Pull the model first: ollama pull modelname. Browse valid model names at Models Hub.