Setup guides
Everything to pick, install and run open models on your own computer.
Open modelsWill it run?Setup guidesLicensesHardwareCheat sheetsProjectsCompare modelsGlossaryRoadmapInterview Q&AFree resources
Setup guides
Step-by-step guides for installing Ollama on Windows, macOS and Linux, plus the basics you need around it: Git, Python virtual environments, what quantization means and how to read a model license. Follow one guide from top to bottom and check the result before you start the next, so any problem is easy to trace to a single step.
Windows
Python
Git
macOS
Linux
Concepts
Common errors and fixes
command not found: ollama
Open a new terminal window after installing. On macOS, open the Ollama app once and accept the command-line link. On Linux check that the install script finished without errors.
The model runs very slowly
Run `ollama ps`. If it shows CPU instead of GPU, update your graphics driver (NVIDIA 551.61 or newer on Windows) and restart Ollama. Otherwise pick a smaller model.
Model downloads fail behind a proxy
Set the HTTPS_PROXY environment variable. The Ollama FAQ advises against setting HTTP_PROXY because it can break downloads.
Other computers cannot reach Ollama
By default Ollama listens only on 127.0.0.1:11434. Set OLLAMA_HOST=0.0.0.0:11434 and restart, only on a network you trust.
The model forgets the start of a long conversation
The default context length is 4096 tokens. Raise it with OLLAMA_CONTEXT_LENGTH or `/set parameter num_ctx 8192` inside a chat. A bigger context uses more memory.
Not enough disk space
Run `ollama list`, then `ollama rm MODEL` for models you no longer use. You can also move storage to another drive with OLLAMA_MODELS.

