Public calculator
GPU to LLM sizing is free to use, no sign-up: VRAM, quantization, single GPU and multi-GPU rigs.
LocalIA is an independent resource published in France. The goal is simple: help freelancers, SMEs and technical teams run open-weight LLMs locally, with readable sizing before you get equipped - without selling you anything.
GPU to LLM sizing is free to use, no sign-up: VRAM, quantization, single GPU and multi-GPU rigs.
Model and GPU pages are built from versioned data and a daily Hugging Face sync.
The site is published by Damien OGRYZEK, Entrepreneur individuel (micro-entreprise), SIREN 918508862.
No sales funnel: you can write to us for advice and get a written reply, with no pushy quote and no commercial follow-up.
The site brings traffic with its free tools. As an independent resource, it mainly has to be clear about its current proof level: no stock, no sales, no bias.
You identify the target model, users, GDPR constraints and approximate budget.
The calculator compares VRAM, CPU, RAM, storage, power and upgrade margin to estimate the right config.
The GPU and LLM comparators and reference configs help you decide with no commercial bias.
You source the parts yourself and install the stack with our guides - or write to us for advice.
The next LocalIA step is independent benchmarks measured on a physical machine. Until then, the site keeps performance as sizing estimates, not promises.
Test the calculator ->