AI Builder · Local-first

Local AI Compatibility Radar v2

Enter your RAM, VRAM and workload to get a practical model-size tier, quantization direction and a cautious performance range. This is a planning aid; verify real speed with your runtime.

No uploadLM Studio / Ollama planningGGUF tier hintsStarter estimates only

Hardware input

Recommended tier

Run the radar.

Expected feel

Speed depends on runtime, drivers, quantization, context and background load.

Best next action

Use this to choose a first test, not as a benchmark claim.

Model fit table

TierExamplesFitNotes

Copyable compatibility report

Run the radar to create a report.

Important limits

Not a live benchmark. The radar uses conservative rules and cannot know your exact drivers, thermal limits or runtime flags.
Do not oversell WebGPU. Browser GPU support varies by OS, browser and hardware. Treat WebGPU as a capability check, not a guarantee.
Use confidence wording. Public recommendations should say starter / proposed / verified instead of absolute claims.