About AICanRun

AICanRun is built by a small group of local AI enthusiasts. We are interested in a simple question: how much useful AI can run directly on the phones people already own?

Running AI locally can mean better privacy, no subscription, and access without an internet connection. But choosing a model is confusing. Phone RAM is not the same as usable memory, model downloads come in many quantizations, and a model that technically fits may still be too slow to enjoy.

What we help you answer

• Will this model fit on my phone?
• How fast might it run?
• Which quant should I download?
• What should I try if it does not fit?

How our results work

Most results begin as estimates based on public device specifications, model file sizes, memory requirements, and memory bandwidth. We do not present an estimate as a real-device test.

• Estimated — calculated from public specifications.
• Measured — backed by reviewed real-hardware evidence.
• ✓ AICanRun verified — tested by us using our published protocol.

See how the calculations and measurement protocol work →

Data and corrections

Device specifications are checked against manufacturer documentation and reputable public sources. Model metadata and file sizes come from public model repositories and official documentation. Device pages show when their specifications were last checked, and unreleased devices are clearly marked as rumored.

Specifications, runtimes, and model files can change. If something looks wrong, email us at hey@aicanrun.com.

Independent by design

AICanRun is an independent project. We are not affiliated with phone manufacturers, model publishers, model-hosting platforms, or local AI app developers. If we add affiliate links or sponsored placements, they will be clearly labeled and will never change a compatibility verdict.