Local models

Local LLM finder

Pick hardware, run server-side WhichLLM, and get practical local model estimates. Browser detection is best-effort; manual selection is the reliable path.

Planning estimates, not visitor-side benchmarks

The website never receives exact VRAM from the browser. When detection is vague, choose the closest manual profile.

Hardware

Detect what the browser exposes, then refine manually.

Hardware profileRecommendation profile

Detected confidencelow

Detected profileManual selection needed

Recommendations

Ready for RTX 4090

idle

Choose hardware and update recommendations

Exact visitor hardware may require manual selection. Results come from server-side WhichLLM, not from running a benchmark in your browser.