Key takeaways
- Install Ollama if you want to run open models locally for privacy, offline work or learning; skip it if you only need the best possible answers.
- On Windows, use the official installer; move models to another drive with the OLLAMA_MODELS environment variable.
- On Linux, the official install script sets Ollama up as a service; on macOS, use the app from ollama.com.
- In Docker, mount a volume for /root/.ollama so models survive container restarts, and pass the GPU through.
- DeepSeek's open models run in Ollama; pick a size that fits your memory.
Install Ollama if you want to run open-weight language models such as Llama, Qwen, Gemma, Mistral or DeepSeek on your own machine, for privacy, offline use, learning or building prototypes without per-token costs. It installs in a few minutes on Windows, macOS and Linux, and runs in Docker. This page helps you decide, then covers installing on Windows including a D drive, where models are stored, macOS, Ubuntu, WSL, Docker and pulling models inside a container, and running DeepSeek models.
Should I install Ollama?
Yes, if you want to learn how language models work hands-on, keep data private, work offline, or prototype apps without paying per token. No, if you only want the best answers for hard tasks, where hosted models such as Claude, ChatGPT or Gemini are much stronger.
Your hardware decides the experience. With 8 GB of RAM you can run small models of a few billion parameters at usable speed. With 16 to 32 GB of RAM or a GPU with 8 GB or more of video memory, mid-size models become practical. Larger models need serious GPUs. Ollama itself is free and open source, and uninstalling it is simple, so trying it costs nothing but disk space: models range from a couple of gigabytes to tens of gigabytes each.
How do I install Ollama on Windows?
Download the Windows installer from ollama.com, run it, and Ollama starts in the background with an icon in the system tray. Then open a terminal and run a model.
Also asked as: how to install ollama on windows
ollama --version
ollama run llama3.2
The first run downloads the model, then opens an interactive chat; type /bye to exit. Ollama also serves an API on http://localhost:11434, which tools such as LangChain and Open WebUI use. If you have an NVIDIA or AMD GPU, keep its drivers updated so Ollama can use it; check with ollama ps while a model runs to see whether it is on the GPU or CPU.
Can I install Ollama on the D drive?
Yes. The models are what take space, and you can store them on any drive by setting the OLLAMA_MODELS environment variable, for example to D:\ollama\models, then restarting Ollama. The installer also accepts a custom install directory.
Also asked as: can i install ollama in d drive
Steps for the models folder:
- Quit Ollama from the system tray.
- Open "Edit environment variables for your account" in Windows settings.
- Add a variable
OLLAMA_MODELSwith the valueD:\ollama\models. - Start Ollama again and pull a model; it downloads to the new folder.
To install the program itself elsewhere, the Windows installer supports a directory option documented in Ollama's Windows guide. Moving existing models is as simple as moving the folder to the new path before restarting.
Where does Ollama install models?
By default, models are stored in a .ollama/models folder in your home directory: C:\Users\<you>\.ollama\models on Windows and ~/.ollama/models on macOS. On Linux, when installed as a service, they are under the ollama user's home, typically /usr/share/ollama/.ollama/models.
Also asked as: where ollama install models
Set OLLAMA_MODELS to change the location on any system. Models are stored as layered blobs with manifests, so the folder will not show friendly filenames; use ollama list to see installed models and ollama rm <name> to delete one and free space. Several models can share layers, so deleting one may free less space than its listed size. If disk space is tight, keep only the models you actually use.
How do I install Ollama on a Mac?
Download the macOS app from ollama.com, move it to Applications and open it. It adds the ollama command to your terminal, and you can run a model straight away with ollama run llama3.2.
Also asked as: how to install ollama on mac
Apple Silicon Macs, with M-series chips, run local models well because the GPU shares fast unified memory with the CPU, so a Mac with 16 GB or more can run mid-size models comfortably. Intel Macs work but are slower. You can also install through Homebrew if you prefer managing tools that way. Keep the app running in the menu bar so the local API at port 11434 stays available for other apps.
How do I install Ollama on Ubuntu?
Run the official install script, which installs Ollama and sets it up as a systemd service that starts automatically.
curl -fsSL https://ollama.com/install.sh | sh
ollama run llama3.2
Check the service with systemctl status ollama and view logs with journalctl -u ollama. For NVIDIA GPUs, install the proprietary drivers first; the script detects them. To change settings such as the models folder or listening address, edit the service with sudo systemctl edit ollama and add environment variables like OLLAMA_MODELS or OLLAMA_HOST, then restart the service. As with any script piped to a shell, you can download and read it first.
Can I install Ollama on WSL?
Yes. Inside WSL 2 on Windows, install Ollama with the Linux install script, and with up-to-date NVIDIA drivers on Windows, Ollama in WSL can use the GPU. Alternatively, install the Windows version and call it from WSL over the local network.
Choose based on your workflow. If your projects live in WSL, running Ollama there keeps everything in one environment. If you also use Windows tools, the Windows app is simpler and you can reach it from WSL at the Windows host's address. Avoid running both at once on the same port. WSL needs enough memory allocated; if models fail to load, raise the memory limit in your WSL configuration.
Should I install Ollama in Docker?
Docker is a good choice for servers, reproducible setups and running Ollama alongside other services such as a web UI or a database. On a personal laptop, the native installer is simpler and usually uses the GPU with less setup.
Run the official image with a volume so models persist:
docker run -d --name ollama -p 11434:11434 -v ollama:/root/.ollama ollama/ollama
For NVIDIA GPUs, install the NVIDIA Container Toolkit and add --gpus=all. With Docker Compose, you can define Ollama, Open WebUI and your app together. Do not expose port 11434 to the internet without authentication in front of it; the Ollama API has no built-in login.
How do I pull Ollama models in a Docker container?
Run the ollama pull command inside the running container with docker exec, and the model is saved in the mounted volume, so it survives restarts.
Also asked as: how to install / pull ollama models in docker container
docker exec -it ollama ollama pull llama3.2
docker exec -it ollama ollama list
docker exec -it ollama ollama run llama3.2 "Explain BM25 in one line"
For automated setups, pull models in a start-up script or call the API's pull endpoint from your app on first run. Keep the volume, named ollama above, when updating the image, so you do not re-download models. If another container needs Ollama, connect them on the same Docker network and use http://ollama:11434 as the address.
Can I install DeepSeek on Ollama?
Yes. DeepSeek's open-weight models, including distilled versions of its reasoning model, are available in Ollama's library. Run one with ollama run deepseek-r1, adding a size tag to choose a smaller or larger variant.
Pick a size that fits your memory: small distilled versions run on ordinary laptops, while the largest DeepSeek models need multiple high-end GPUs and are not practical locally. Reasoning models print their thinking before the final answer, so expect longer responses. Running locally means your prompts stay on your machine, which addresses privacy concerns some people have about using hosted services. Test on your own tasks; smaller distilled models are much weaker than the full model.
Sources
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. For campus sessions the inviting institution covers the visit end to end: travel (a First AC train ticket for journeys up to five hours, a flight for anything longer), accommodation, meals, and pickup and drop; the honorarium and workshop fees are negotiable on WhatsApp or on the call, and requests are welcome from institutions that can cover the visit this way. Email pranjulrathour41@gmail.com.
