Ollama Local LLMs Cheat Sheet

Installation

# macOS
curl -fsSL https://ollama.com/install.sh | sh

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# Windows
# Download installer from ollama.com

# Verify installation
ollama --version

# Start Ollama server
ollama serve

Basic Commands

Popular Models

Model Variants

# Pull specific quantization
ollama pull llama3.3:8b-q4_0          # 4-bit quantization
ollama pull llama3.3:8b-q8_0          # 8-bit quantization
ollama pull llama3.3:70b-q4_K_M       # Medium quality 4-bit

# Quantization levels (smallest to largest):
# q2_K - Very small, lowest quality
# q3_K_S, q3_K_M, q3_K_L - Small, low quality
# q4_0, q4_1, q4_K_S, q4_K_M - Medium size/quality (recommended)
# q5_0, q5_1, q5_K_S, q5_K_M - Larger, better quality
# q6_K - Very large, high quality
# q8_0 - Largest, highest quality
# f16 - Full precision (huge)

Interactive Mode

# Start chat
ollama run llama3.3

# Special commands in chat:
/bye           # Exit
/clear         # Clear conversation
/set parameter # Set parameters
/show info     # Show model info
/? or /help    # Show help

# Examples:
>>> /set temperature 0.7
>>> /set top_p 0.9
>>> /show info
>>> /clear

API Usage

# Python
pip install ollama

import ollama

# Simple generation
response = ollama.chat(model='llama3.3', messages=[
    {
        'role': 'user',
        'content': 'Why is the sky blue?',
    },
])
print(response['message']['content'])

# Streaming
for chunk in ollama.chat(
    model='llama3.3',
    messages=[{'role': 'user', 'content': 'Tell me a story'}],
    stream=True,
):
    print(chunk['message']['content'], end='', flush=True)

# With options
response = ollama.chat(
    model='llama3.3',
    messages=[{'role': 'user', 'content': 'Hello'}],
    options={
        'temperature': 0.7,
        'top_p': 0.9,
        'top_k': 40,
    }
)

REST API

# Generate completion
curl http://localhost:11434/api/generate -d '{
  "model": "llama3.3",
  "prompt": "Why is the sky blue?",
  "stream": false
}'

# Chat completion
curl http://localhost:11434/api/chat -d '{
  "model": "llama3.3",
  "messages": [
    {"role": "user", "content": "Hello!"}
  ]
}'

# List models
curl http://localhost:11434/api/tags

# Show model info
curl http://localhost:11434/api/show -d '{
  "name": "llama3.3"
}'

# Pull model
curl http://localhost:11434/api/pull -d '{
  "name": "mistral"
}'

Model Options

Custom Modelfile

# Create Modelfile
FROM llama3.3

# Set parameters
PARAMETER temperature 0.7
PARAMETER top_p 0.9
PARAMETER top_k 40

# Set system prompt
SYSTEM """
You are a helpful coding assistant specializing in Python.
Provide clear, well-commented code examples.
"""

# Build model
ollama create my-python-assistant -f ./Modelfile

# Use custom model
ollama run my-python-assistant

Embeddings

# Python
import ollama

# Generate embeddings
embedding = ollama.embeddings(
    model='llama3.3',
    prompt='The sky is blue'
)
print(embedding['embedding'])

# REST API
curl http://localhost:11434/api/embeddings -d '{
  "model": "llama3.3",
  "prompt": "The sky is blue"
}'

# Use nomic-embed-text for embeddings (recommended)
ollama pull nomic-embed-text
embedding = ollama.embeddings(
    model='nomic-embed-text',
    prompt='Your text here'
)

LangChain Integration

from langchain_community.llms import Ollama
from langchain.callbacks.manager import CallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler

llm = Ollama(
    model="llama3.3",
    callback_manager=CallbackManager([StreamingStdOutCallbackHandler()])
)

response = llm.invoke("Tell me a joke")

# With chat models
from langchain_community.chat_models import ChatOllama

chat = ChatOllama(model="llama3.3")
messages = [
    ("system", "You are a helpful assistant"),
    ("human", "What is 2+2?")
]
response = chat.invoke(messages)

Performance Optimization

Environment Variables

# Set model storage location
export OLLAMA_MODELS=/path/to/models

# Set host and port
export OLLAMA_HOST=0.0.0.0:11434

# GPU settings
export OLLAMA_NUM_GPU=1        # Number of GPUs to use
export OLLAMA_GPU_LAYERS=35    # Number of layers to offload

# Keep models loaded
export OLLAMA_KEEP_ALIVE=5m    # Keep in memory for 5 minutes

Troubleshooting