Ollama (local AI): hardware requirements
Run LLMs locally — chat, code completion, private AI.
Hardware tiers
| Tier | CPU | RAM | Storage | GPU | VRAM | Network |
|---|---|---|---|---|---|---|
| Minimum sensible | 8 cores | 16 GB | 1 TB (ssd) | ai | 12 GB VRAM | 1 GbE |
| Recommended | 12 cores | 32 GB | 1 TB (ssd) | ai | 16 GB VRAM | 2.5 GbE |
| Comfortable | 16 cores | 64 GB | 2 TB (ssd) | ai | 24 GB VRAM | 10 GbE |
| Overkill | 26 cores | 103 GB | 4 TB (ssd) | ai | 24 GB VRAM | 10 GbE |
- Minimum sensible: The smallest build that genuinely works — tight, focused, no fluff.
- Recommended: The sweet spot: handles the service plus the usual side-services.
- Comfortable: Real headroom — this service will never be the bottleneck.
- Overkill: More than you will ever use for this service alone — spend the money elsewhere.
Beyond Comfortable — 16 cores, 64 GB RAM, 2 TB: headroom you will never use for this service alone.
Size it exactly
Pick the model you want to run — the VRAM need is computed, not guessed.
You need about 8 GB of VRAM for this.
Runs well with
Image generation, AI experimentation, Databases
Frequently asked questions
How much RAM does Ollama (local AI) need?
16 GB is the sensible minimum, 32 GB covers most real setups, and 64 GB gives comfortable headroom.
How many CPU cores does Ollama (local AI) need?
8 cores are the minimum, 12 cores the sweet spot, and 16 cores comfortable when shared with other services.
Can I run several services on one machine?
Yes — most services share resources. Use the planner and select several services; the engine sums and overlaps the requirements instead of stacking them.
What if I already have hardware?
Register it in My hardware and the analysis shows exactly which services your current system can comfortably run — including an honest “do nothing”.