What this estimate does
It estimates model weights, context memory and runtime workspace, then checks whether each Q4 size band fits fully in GPU memory, needs hybrid GPU/RAM offload, falls back to CPU/system RAM, or does not fit safely.
FREE · PRIVATE · BROWSER-ONLY
Can My PC Run Local AI?
Enter RAM, VRAM, workload and context for a conservative model-fit shortlist, including smaller CPU-only PCs. Everything is calculated in this page. Nothing is uploaded and this is not a benchmark or compatibility guarantee.
0 trackers0 cookies0 API calls0 hardware uploads
Your hardware
Try an example, not a hardware scan:
Integrated/shared GPU memory is not the same as dedicated VRAM. If you have no dedicated GPU, enter 0.
Where do I find RAM and VRAM on Windows?
- Press Ctrl + Shift + Esc to open Task Manager.
- Choose Performance → Memory for installed RAM.
- Choose Performance → GPU and find Dedicated GPU memory. If multiple GPUs appear, select the device your AI runtime will use.
Shared GPU memory is not dedicated VRAM. This estimator reduces raw totals to conservative usable budgets.
ESTIMATE
Ready for a first pass
Use the form to see a conservative hardware-fit range.
Recommended starting point—
Fit mode—
Estimated working set—
See where the memory goes
One shared scale for estimated model working memory. Fit modes still depend on GPU versus RAM placement.
Estimated allocation, not a live PC scan, speed measurement, or compatibility guarantee.
Save or share this estimate
Nothing is uploaded. Your browser copies only the text you choose.
What “fits” does not mean
A model fitting in memory does not prove that it will be fast or even supported by your runtime. GPU architecture, memory bandwidth, exact quantisation and model implementation still decide real performance.
Next: verify with the real machine
LLMRadar is built around hardware-aware model recommendations and local runtime evidence. HardwareRadar focuses on the Windows hardware and telemetry layer.