Ollama CLI Cheat Sheet: Every Command You Need

Ollama is an open-source tool that lets you download, run, and manage large language models locally on Linux, macOS, and Windows. This cheat sheet covers every CLI command, Modelfile instruction, REST API endpoint, and environment variable you are likely to need — structured as a dense reference you can bookmark and return to rather than read top to bottom.

Installation Quick Reference

Install

Verify and start

Default API port: 11434
API base URL: http://localhost:11434

Model Management

Understanding model tag conventions

Model tags follow the pattern <family>:<size>-<type>-<quantisation>. For example, llama3.2:8b-instruct-q4_K_M breaks down as:

If you omit the tag (e.g. ollama pull llama3.2), Ollama downloads the default recommended tag for that model.

Running Models

ollama run flags

Interactive session slash commands

These commands are available once you are inside an ollama run session.

Modelfile Reference

A Modelfile is a plain-text configuration file (similar in concept to a Dockerfile) that defines a custom model — its base, system prompt, parameters, and template. You create a custom model by pointing ollama create at a Modelfile.

Modelfile instructions

Minimal working Modelfile example

FROM llama3.2
PARAMETER temperature 0.7
PARAMETER num_ctx 4096
SYSTEM "You are a helpful assistant."

Building a model from a Modelfile

Key Parameters Reference

REST API Quick Reference

Endpoints

curl examples

Generate (non-streaming):

curl http://localhost:11434/api/generate \
  -d '{"model":"llama3.2","prompt":"What is Linux?","stream":false}'

Chat (non-streaming):

curl http://localhost:11434/api/chat \
  -d '{
    "model": "llama3.2",
    "messages": [{"role": "user", "content": "What is Linux?"}],
    "stream": false
  }'

Environment Variables

Service Management

Linux (systemd)

Setting environment variables on Linux (systemd)

Add Environment lines under the [Service] section of /etc/systemd/system/ollama.service:

[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"
Environment="OLLAMA_KEEP_ALIVE=10m"

Then reload the daemon and restart the service:

systemctl daemon-reload && systemctl restart ollama

macOS

Model Storage Locations

Directory structure

Moving models to a different disk: Set OLLAMA_MODELS to the new path, move the models/ directory to that location, then restart the Ollama service.

Useful One-Liners

Check total disk space used by models:

du -sh ~/.ollama/models/

List all local models with sizes:

ollama list

Unload a specific model from memory immediately:

curl -X POST http://localhost:11434/api/generate \
  -d '{"model":"llama3.2","keep_alive":0}'

Test that the API is responding:

curl http://localhost:11434/api/tags

Run a model with an inline system prompt:

ollama run llama3.2 --system "You are a Linux expert." "Explain systemd"

Generate embeddings for a string:

curl http://localhost:11434/api/embed \
  -d '{"model":"nomic-embed-text","input":"Hello world"}'

Pull multiple models in parallel (bash):

ollama pull llama3.2 & ollama pull nomic-embed-text & wait

Show which models are currently loaded in VRAM/RAM:

curl http://localhost:11434/api/ps

Further guides available on serverman.co.uk cover related topics in depth: securing Ollama for network exposure (the Ollama security guide), integrating Ollama as an AI backend in automation workflows (how to use Ollama with n8n), connecting local models to agentic toolchains (building local AI agents with Ollama and MCP), and using Ollama as a drop-in replacement for the OpenAI API in existing applications (the Ollama OpenAI API compatibility guide).