- The official and most effective method for updating is to run the OllamaSetup.exe installer again.
- Model storage paths can be managed using the OLLAMA_MODELS environment variable.
- The tool is compatible with NVIDIA and AMD GPUs, requiring updated drivers for optimal performance.
- It allows the local execution of models such as Llama 3 and the integration of external visual interfaces.
If you're into the world of artificial intelligence, you probably already know Ollama. It's that open-source tool that lets you build language models on your own computer without having to rely on the cloud. Best of all, it's super versatile and lets you run powerful LLMs without your PC needing to be a NASA supercomputer, although of course, if you have more modest hardware, token generation will be a bit slower.
Installing it is a piece of cake, but sometimes you wonder how to keep the software up to date or how to fix those common errors when you try to update. To save you from getting bogged down in technical trouble, we've prepared this complete guide where we'll tell you everything you need to know. Everything you need to know to make Ollama fly on your Windows system.
How to update Ollama on Windows correctly
Many people make the mistake of trying to update Ollama using Linux commands, such as the famous curl shwithin the Windows terminal. If you do this, you may see messages saying that the previous version is being cleaned up, but when you restart the program, the version remains the sameThis happens because the Windows installer doesn't work using shell scripts in that way.
The simplest and most effective way to stay up to date is to run the file again. OllamaSetup.exeThis installer is smart and handles the binaries update without deleting your settings or previously downloaded models. Essentially, it's the fast track to getting the latest improvements and support for the newest models.
If you prefer a more manual installation or need to integrate Ollama as a system service, there's a .zip file (ollama-windows-amd64.zip) that contains only the CLI and dependencies. In this case, it's essential if you're upgrading. delete old directories before decompressing the new ones to avoid file conflicts.
System requirements and hardware optimization
For everything to run smoothly, you need Windows 10 (version 22H2 or later), either Home or Pro. Regarding graphics power, Ollama is compatible with NVIDIA and AMD Radeon. If you have an NVIDIA card, make sure you have the drivers 551.61 or higherFor AMD users, the ROCm v7 or HIP7 driver stack is recommended, although the system usually resorts to Vulkan as an alternative if the hardware is not fully compatible.
Note this curious detail: on some older Windows 10 terminals, Ollama's progress characters may appear as strange little squares This is due to Unicode. If this happens to you, don't panic, simply change your terminal's font to a more modern one and the problem will disappear.
File management and model location
By default, Ollama installs in your user folder and does not request administrator privileges. However, language models may occupy tens or hundreds of gigabytesAnd it's very likely that your C: drive will quickly become too small. To prevent the system from crashing, you can change the path where these files are saved.
To move the models, you must create an environment variable called OVEN_MODELS In your user account settings, simply go to Control Panel or Settings, find Environment Variables, and assign the path to the folder where you want to store the LLM files. After doing this, remember to close the application from the system tray and reopen it for the changes to take effect.
If you need to investigate what's happening under the hood, you can access the logs using the command explorer %LOCALAPPDATA%\OllamaThere you will find the upgrade.log To see how the updates have gone, use app.log for the graphical interface and server.log for server processes.
Use of models and deployment of interfaces
Once updated, you can start playing with models like Llama 3, Phi 3, Mistral, or Gemma. To download and run one, simply type the following into the console. ollama run llama3One important piece of advice is leave twice the free space on disk compared to the weight of the model you want to install to avoid write errors.
If the command console seems too basic to you, you can install Open WebUI using Docker. This allows you to have a visual interface very similar to ChatGPT's, making interaction with your local models much more intuitive and visually appealing, and even learning. How to share Open WebUI with other users of your network.
Furthermore, Ollamama gives you the freedom to create your own LLM Customized. This means you can configure the system to combine different models or adjust the parameters so that the AI behaves exactly as you want, always maintaining your privacy by not sending data to external servers.
Having Ollama updated on Windows ensures compatibility with the latest GPUs, optimizes response speed, and allows you to manage the location of heavy models using environment variables, all while enjoying a flexible interface whether through commands or external tools like Docker.
I am a technology enthusiast who has turned his "geek" interests into a profession. I have spent more than 10 years of my life using cutting-edge technology and tinkering with all kinds of programs out of pure curiosity. Now I have specialized in computer technology and video games. This is because for more than 5 years I have been writing for various websites on technology and video games, creating articles that seek to give you the information you need in a language that is understandable to everyone.
If you have any questions, my knowledge ranges from everything related to the Windows operating system as well as Android for mobile phones. And my commitment is to you, I am always willing to spend a few minutes and help you resolve any questions you may have in this internet world.