AI/ML

Host Ollama on VeloxaHost

Ollama downloads and runs open-weight large language models such as Llama, Mistral and Gemma on your own server, with a command-line tool and a REST API on port 11434. Developers self-host it to keep prompts and data off third-party APIs and to give internal tools a private model endpoint.

The one-click install runs Ollama's official install script on Ubuntu 24.04, enables the ollama systemd service and pulls llama3.2:1b, a 1.3 GB model, so there is something to test straight away.

Install it yourself

  1. Run the official install script. It installs the binary, creates an ollama user and sets up the systemd service.
    curl -fsSL https://ollama.com/install.sh | sh
  2. Check that the service is running.
    systemctl status ollama
  3. Pull a model. llama3.1:8b is a 4.9 GB download.
    ollama pull llama3.1:8b
  4. Try it from the command line.
    ollama run llama3.1:8b "Explain DNS in two sentences."
  5. Call the REST API, which listens on 127.0.0.1:11434 by default.
    curl http://localhost:11434/api/generate -d '{"model": "llama3.1:8b", "prompt": "Hello", "stream": false}'
  6. Only if other machines need the API: run systemctl edit ollama, add the lines below, then systemctl restart ollama. Ollama has no authentication of its own, so restrict who can reach the port.
    [Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"

FAQ

What license is Ollama under?

Ollama is released under the MIT license. The models you pull have their own licenses, set by whoever published them, so check each model's terms before using it in a product.

Does Ollama get a GPU on VeloxaHost?

No. VeloxaHost's tenant projects block GPU passthrough, so Ollama runs models on the CPU. That suits small models, batch jobs and internal tools, but expect much slower generation than on a GPU.

How much memory do I need for a model?

The model has to fit in RAM along with its context, so memory should comfortably exceed the model's download size. As reference points, llama3.2:1b is 1.3 GB and llama3.1:8b is 4.9 GB. The recommended 8 GB of RAM leaves room for one 7B or 8B model.

Where are models stored, and how do I update Ollama?

On Linux, models are kept in /usr/share/ollama/.ollama/models; set OLLAMA_MODELS to move them. To update Ollama, run the install script again: curl -fsSL https://ollama.com/install.sh | sh.

Can I try Ollama on the free trial?

Not at its minimum size. Ollama needs at least 4 vCPU, 4 GB of RAM and 30 GB of disk, and the trial's allowance is 2 vCPU, 4 GB of RAM and 40 GB of disk. Hourly VPS sizes need billing activated, even during the trial.

Which VeloxaHost size fits Ollama?

The recommended 4 vCPU and 8 GB of RAM is a Veloxa 4: 4 vCPU, 8 GB RAM and 160 GB of disk for $40 a month, billed at $0.06 an hour and capped at the monthly price. CPU and memory on a container can be raised from the console while it runs, if you move to a larger model.