SEODesktop

Ollama (local models)

Ollama runs large language models locally on your machine. There are no API calls to an external service, no per-token cost, and nothing leaves your network. The trade-off is speed — local models are slower than cloud APIs, and quality depends on the model and your hardware.

Ollama is the best pick when privacy matters most, or when you want to avoid per-query cost entirely.

Install Ollama

  1. Download the installer from ollama.com/download — Windows, macOS and Linux builds are available.

  2. Install and run it. On first launch it starts a local HTTP server on http://localhost:11434.

  3. Pull a model from a terminal:

    ollama pull llama3.3

    Smaller models download in seconds; larger ones (70B) can be several gigabytes.

Recommended models for SEO work

ModelSizeBest for
llama3.3~40GBGeneral-purpose analysis; strong reasoning.
llama3.1:8b~5GBFast bulk classification on low-RAM machines.
qwen2.5:14b~9GBGood balance of quality and speed.
mistral:7b~4GBLightweight; fine for summaries.

Larger models give better analysis but need more RAM — a rough guide is model size × 1.3 = required free RAM.

Configure in SEODesktop

  1. Open Settings → AI Providers → Ollama.
  2. Leave the endpoint at the default http://localhost:11434 unless you've configured Ollama to run on a different host/port.
  3. The dropdown will list every model you've pulled. Pick a default.
  4. Click Test connection. A successful test generates a short response using the selected model.
  5. Click Save.

Performance tips

  • Apple Silicon and modern GPUs give a big speedup. On CPU-only, expect 5–20 tokens/sec — fine for bulk classification, too slow for interactive use.
  • Keep the Ollama app running in the background; SEODesktop talks to it over HTTP.
  • Pre-warm the model by running ollama run <model> once before you start a large batch — the first call after idle takes a few seconds to load the weights into memory.

Privacy

Nothing leaves your machine. All prompts, responses and model weights live on local disk. SEODesktop's Ollama integration uses the standard Ollama HTTP API — no telemetry is added by SEODesktop.