CHECK YOUR LOCAL AI MEMORY USAGE
SEE HOW YOUR MODEL USES YOUR SYSTEM
Learn how to inspect a running model in Ollama and monitor memory usage in Windows. This guide helps you check what is loaded, understand GPU and system memory use, and compare what changes when you run or stop a model.
BEFORE YOU BEGIN
START WITH A WORKING LOCAL MODEL
This guide follows the Windows and Ollama setup used in Run Your First Local AI Model. Complete that walkthrough first so you have Ollama installed and a local model ready to run.
Keep PowerShell and Windows Task Manager available during the checks. Task Manager lets you monitor system memory and GPU performance while you compare your computer before, during, and after running a model.
CHECK YOUR STARTING POINT
OPEN WINDOWS TASK MANAGER
Press Ctrl + Shift + Esc to open Task Manager. Select Performance, then Memory. If you see a compact view without tabs, select More details first.
Before starting your model, note the memory shown as In use and Available. These readings give you a starting point for comparison. Windows, background services, and other applications already use some memory, so the starting usage will not be zero.
Leave Task Manager open for the next steps. Keep the same applications open during your comparison so changes are easier to interpret.
RUN YOUR LOCAL MODEL
LOAD THE SAME MODEL YOU TESTED BEFORE
Open PowerShell and enter:
ollama run gemma3:1b
This starts the small local model used in the first walkthrough. If it is no longer downloaded, Ollama will download it before starting.
When the chat prompt appears, enter:
Explain system RAM in two short sentences.
Wait for the response, then leave the chat open. You will inspect the loaded model from a second PowerShell window in the next step.
COMPARE MEMORY USAGE
CHECK WINDOWS WHILE THE MODEL IS LOADED
Return to Task Manager → Performance → Memory. Compare In use and Available with the starting values you noted earlier. These readings cover your whole computer, so any change may include activity from Windows and other applications.
If Task Manager lists a GPU under Performance, select it and look for Dedicated GPU memory usage and Shared GPU memory usage. If you have multiple GPUs, check each one to find where memory usage changed.
On a discrete graphics card, dedicated GPU memory is the card’s VRAM. Shared GPU memory uses system RAM; it is not extra VRAM installed on the card. Integrated graphics use a different memory arrangement, so their readings may look different.
Use these readings alongside ollama ps: Ollama reports model placement, while Task Manager shows overall memory usage.
USE WHAT YOU LEARNED
COMPARE ONE CHANGE AT A TIME
Record the model name, its PROCESSOR result from ollama ps, and the memory readings before loading, while loaded, and after stopping. These notes give you a useful baseline for future comparisons.
When trying another model, stop the previous one first and repeat the same checks with the same short prompt and applications open. Change one thing at a time so the results are easier to interpret.
A successful test with this small model does not establish how much memory a larger model will need. Model choice, context settings, and simultaneous requests can change memory requirements. Repeat the checks with the setup you actually plan to use.
SOURCES & VERIFICATION
CHECK THE DOCUMENTATION
This guide uses Ollama’s official command-line documentation and Microsoft’s explanations of Windows Task Manager and GPU memory.
Ollama — CLI Reference
Ollama — FAQ
Microsoft — System Configuration Tools in Windows
Microsoft — GPUs in Task Manager
Last checked against official documentation: October 19, 2026.