Updated —
Checked every week
★ GitHub
🖥️ Local AI Only · No Cloud · No API Keys

Will your LLM run
on this machine — locally?

Pick any model, configure your hardware, and see exactly where the memory goes — in seconds.

✅ Only models you can download & run on your own hardware — no subscriptions, no internet required
🤖
nerds visited
visitor count
and counting 🚀
65 models · list checked every Monday — new sizes of tracked families are added automatically
Your device
Hardware
GPU—
RAM tier—
Platform—
or enter manually
AI company
Model
Quantization
?
What is quantization?
Think of it like image quality. Higher = sharper but bigger file. Lower = smaller file, slight quality drop. It controls how precisely the AI model's "knowledge" is stored — and directly determines how much RAM it needs.
FP16Full quality — needs the most RAM. Best results.
Q8Near-perfect quality, half the size of FP16.
Q4Sweet spot — great quality, runs on most laptops.
Q3Smaller, but noticeably less accurate.
Q2Tiny file, significant quality loss. Not recommended.
Q4_K_M: ~4.8 bits/weight — Ollama default, best quality-size balance.
Context length
4K
🔍
Configure your hardware
Select brand, chip, and RAM on the left — then pick a model to see full memory breakdown.
💬 Feedback / Suggestions