About AI Local Check

AI Local Check answers one question as accurately as we can: can your computer run this AI model locally? You give it a model and your hardware; it tells you which GGUF quantizations fit your VRAM and RAM, how much memory each needs, and whether the model runs fully on your GPU, only with offloading, or not at all.

Our one rule: the numbers must be real

There is a lot of vague, copied or outdated "will it run" advice online. We built this site because we wanted an answer we could actually trust. So every figure here is computed from a model's real published files and configuration — the actual GGUF file sizes, the parameter count from its metadata, and the KV cache from its true architecture. We never invent numbers, and we never use a language model to fabricate facts. If we can't ground a claim in real data, we don't publish it.

You can read exactly how the calculations work on our methodology page.

What you'll find here

• A memory breakdown for thousands of models, by quantization, on the home page.
• Per-GPU pages showing what each graphics card can run, under By GPU.
• In-depth guides on VRAM, quantization, hardware choice and running specific models locally.
• Measured, data-backed coverage of new model releases in News.

What we don't do

We don't publish fabricated benchmarks, invented specifications, or hype. Speed figures are clearly marked as estimates, because real throughput depends on your exact setup. When a model's architecture can't be determined from public data, we say so instead of guessing.

Who is behind it

AI Local Check is published by Genisoft France, a company registered in Echirolles, France (SIRET 815 108 352 00023), with Lefi Abdelmonem as publication director. Full publisher and legal details are on our contact & legal notice page.

Found a number that looks off, or want to suggest a model or GPU to add? We genuinely want the feedback — email contact@genisoft.fr.