Local AI hardware checker
Which AI models
can your PC run?
Check GPU memory, RAM and storage before you download a model. See what fits, what needs a compromise, and why.
116 completed hardware scans
What this count means
Each completed detector scan adds one. Repeat scans and early development tests are included, so this is not a count of unique people. Page views, manual entries and example PCs do not add to it. The total survives deletion of temporary hardware profiles. This page refreshes the count periodically.
Know your GPU? Get a quick estimate
Have a model in mind? Use the LLM VRAM calculator.
Example PC · Hypothetical configuration
What runs on an RTX 4070?
4,096 token context · Ollama
- Graphics
- GeForce RTX 4070
- Dedicated VRAM
- 12 GB
- System RAM
- 32 GB
- Processor
- Ryzen 7 7800X3D
Example assumptions
Windows 11; 620 GB free storage; CUDA available. RAM available: 21 GB. GPU budget: 11 GB; currently free: 9 GB. These values are illustrative. The driver version is unknown. The engine applies its own memory reserves. Runtime-version compatibility is not checked in this estimate; use an Ollama release that supports the selected build.
| Model / exact build | Download | Est. memory | Result on this example PC |
|---|---|---|---|
| llama3.2 llama3.2:1b-text-q4_K_M Est. memory: 1.4 GB Download: 0.8 GB | 0.8 GB | 1.4 GB | Excellent fit medium confidence |
How this result was calculated
Assumptions and missing information
GB uses 1,024³ bytes. Memory includes weights, context cache and runtime overhead. | |||
| llama3.1 llama3.1:8b-text-q4_K_M Est. memory: 5.8 GB Download: 4.6 GB | 4.6 GB | 5.8 GB | Excellent fit medium confidence |
How this result was calculated
Assumptions and missing information
GB uses 1,024³ bytes. Memory includes weights, context cache and runtime overhead. | |||
| gpt-oss gpt-oss:20b Est. memory: 14.5 GB Download: 12.8 GB | 12.8 GB | 14.5 GB | Runs with trade-offs medium confidence |
How this result was calculated
Assumptions and missing information
GB uses 1,024³ bytes. Memory includes weights, context cache and runtime overhead. | |||
| llama3.3 llama3.3:70b Est. memory: 44.4 GB Download: 39.6 GB | 39.6 GB | 44.4 GB | Not recommended medium confidence |
How this result was calculated
Assumptions and missing information
GB uses 1,024³ bytes. Memory includes weights, context cache and runtime overhead. | |||
Estimates for this example PC. Missing model or driver information lowers confidence. Expand a result for the memory checks and assumptions. No inference benchmark was run.
Explore the full exampleLooking for another model?
Compare model sizes and builds, or start with the GPU you own.
Before you check
What the tool needs, and what a compatibility result can tell you.
Read the local AI setup guideDo I need to download anything?
Use the reference estimate above, or enter your full specs manually. For automatic detection, Windows users download and double-click the executable. Mac and Linux users run the Python detector in Terminal with Python 3.9 or newer. The scan page gives you the command.
Does “fits” mean it will be fast?
A fit means the engine estimates enough memory and a compatible execution path at the context shown. Speed also depends on your processor, memory bandwidth, runtime and settings. These results do not include measured tokens per second.
Why is memory use larger than the download?
Running a model also needs a cache for the conversation and memory for the runtime. Longer conversations increase that cache. The results page lets you change the context length and recalculate.
What does the detector read?
It reads CPU, RAM, GPU memory, acceleration support and free disk space, then sends a hardware profile to your temporary scan. It does not read your documents or browser history. Read the privacy policy.