Ollama: Run AI Models for Free on Your Own Server

What Is Ollama?

Ollama is an open-source tool that makes it trivially easy to download and run large language models (LLMs) locally. With a single terminal command, you can have a capable AI model running on your own hardware — no OpenAI account required, no per-token billing, no data leaving your machine.

It supports dozens of models including Llama 3, Mistral, Gemma, Phi, DeepSeek, Qwen, and many others. It also exposes a local REST API that's compatible with the OpenAI API format — meaning tools built for OpenAI can be pointed at Ollama with minimal changes.

Why Run AI Locally?

The business case for Ollama comes down to three things:

• Cost: OpenAI, Anthropic, and other API providers charge per token. At scale — thousands of automated content generations per day — this adds up fast. Ollama costs nothing beyond your server bill.
• Privacy: Your prompts and outputs never leave your infrastructure. Critical for business-sensitive automation workflows.
• Control: No rate limits, no terms of service changes that break your automation, no dependency on third-party uptime.

Installing Ollama

On Linux (including a VPS), installation is one command:

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

On Mac, download the app from ollama.com. On Windows, use the installer or WSL2.

After installation, pull your first model:

ollama pull llama3.2

Then run it:

ollama run llama3.2

That's it. You're now running a capable AI model entirely on your own hardware.

Recommended Models for Different Use Cases

For content generation (scripts, posts, emails):

• llama3.2 — fast, good quality, runs on modest hardware
• mistral — excellent instruction following, great for structured output
• qwen2.5 — outstanding for business and technical writing

For coding and technical tasks:

• deepseek-coder-v2 — one of the best open-source coding models available
• codellama — Meta's code-focused model, reliable for Python and JavaScript

For lightweight/fast tasks on low-spec hardware:

• phi3 — Microsoft's compact model, excellent quality-to-size ratio
• gemma2:2b — Google's 2B model, very fast, surprisingly capable

Using Ollama in Automation Workflows

Ollama's local API runs on http://localhost:11434 and accepts the same request format as OpenAI. This means you can replace OpenAI API calls in your n8n workflows, Python scripts, or any other tool with Ollama calls at zero cost.

Example n8n HTTP Request node config for Ollama:

• URL: http://localhost:11434/api/generate
• Method: POST
• Body: {"model":"llama3.2","prompt":"Write a short YouTube script about passive income","stream":false}

This generates AI content in your automation pipeline without a single API token being consumed.

Ollama on a VPS: The Ideal Setup

Running Ollama on a dedicated VPS (rather than your laptop) means your automation workflows can call it 24/7 without your computer needing to be on. A VPS with 8GB RAM and a modern CPU handles 7B parameter models comfortably. GPU-enabled VPS servers handle 13B–70B models for higher quality output.

The cost: a 4-core, 8GB VPS typically runs £8–£20/month depending on the provider — often less than a single month of moderate OpenAI API usage at automation scale.

Practical Use Cases for Income Automation

• Generate 50 social media posts daily at zero cost
• Create product descriptions, email sequences, and ad copy in bulk
• Summarise and repurpose long-form content into short clips or posts
• Generate SEO-optimised blog drafts for review and publishing
• Power a customer-facing chatbot with no per-message fees
• Run private, sensitive business workflows without data leaving your infrastructure

Ollama vs Paid APIs: When to Use Each

Ollama is ideal for high-volume, cost-sensitive automation tasks and private data. Paid APIs (OpenAI, Anthropic) still have the edge for the most demanding tasks — very long context, frontier reasoning, or the highest quality output. A smart setup uses Ollama for 80% of tasks and reserves paid API calls for critical, high-value outputs.