Model Atlas™

The right model.
For this machine.

Model Atlas matches AI models to your hardware and your use case — so you can do more, locally.

MORE
CAPABILITY
A BRIGHTER
TOMORROW.

Your machine

Choose a sample profile. Customize your exact RAM and GPU memory below.

Customize hardware and use case

Enter physical memory, not storage. Integrated graphics use the CPU estimate here; if your PC has a dedicated GPU, enter its GPU memory. Specs stay in this browser tab.

What are you building?

Your match

Recommended

Explore models

Comfortable Possible Unlikely locally

LOCAL DEVICE READOUT

What this browser can see.

Browser hints can be masked or rounded. Confirm your specs for a personal match.

Scanning this browser…

Not exposed to websitesExact chip, installed RAM, GPU memory, disk capacity, and laptop versus desktop form factor. Confirm these specifications before relying on a match.

AI CAPABILITY MAP

What this machine can do.

Concrete local workloads mapped to a representative model and this setup’s estimated memory budget.

Sample profile

Memory fit is a planning estimate. Runtime support and speed are not measured.

Browser hints and entered specs stay in this tab. The public page cannot verify exact RAM or GPU memory. Model fit uses illustrative Q4-class estimates; no model runs.

How these matches work

Model Atlas compares an illustrative quantized loading footprint with your entered GPU memory or a reserved portion of system memory. A comfortable fit has headroom under that rule; possible can mean CPU offload or a tight shared-memory fit; unlikely exceeds the demo allowance.

These are planning estimates, not measured benchmarks or official requirements. Context length, vision inputs, quantization format, runtimes, drivers, and other apps can change fit and speed. Task suitability is curated from publisher material, not a measured quality score.

Publisher model cards are linked in each model insight. Some models require special runtimes or licenses. Apple Foundation Models is a native platform API, separate from downloadable OpenELM and FastVLM weights; device and OS eligibility must be checked on the Apple platform. Explore Apple’s framework ↗

Compare models

The same hardware and use case, side by side.