LIBRARY / TOOLS AND REFERENCES

Knowledge you
can verify.

Guides for sizing on-premises AI, comparing alternatives and understanding the calculator’s results.

ARTICLES / EFRÉN FIGUEROA

Judgement for your infrastructure.

Perspective

Beyond silicon: judgement and method for bringing AI to your infrastructure

A practical bridge between AI models and infrastructure: memory, performance, sizing and total cost, guided by technical judgement.

Read article ↗

GUIDES / EFRÉN FIGUEROA

From model to deployment.

Videos in Spanish

How to use the Calculator

A sequence of 15 videos explaining the LLM Calculator. Original videos in Spanish.

Read guide ↗

Sizing

How much VRAM does an LLM need? A calculation guide

Learn to calculate the VRAM an LLM needs from its parameters, precision and KV cache. Formulas, examples and a calculator for your own scenario.

Read guide ↗

Sizing

LLM quantization: FP16 vs INT8 vs INT4

What LLM quantization is, how it affects VRAM and performance, and how to choose between FP16, INT8 and INT4 for your use case.

Read guide ↗

Sizing

What is the KV cache in an LLM, and why does it consume so much memory?

What the KV cache is, how it grows with context and concurrency, and how to estimate its memory requirements when sizing a server.

Read guide ↗

Economics

On-premises vs cloud AI for businesses: when each makes sense

Compare costs, data control and operations for on-premises LLMs and cloud APIs. Make a decision based on your actual use case.

Read guide ↗

Sizing

Single server vs AI cluster: when to scale out

When one GPU server is enough for on-premises AI and when a cluster makes sense. Concurrency, availability and cost criteria.

Read guide ↗

AVAILABLE NOW

TOOL

LLM Calculator

Configure hardware, adjust the workload and download the report.

Open resource ↗

METHODOLOGY

Memory and performance

Formulas, assumptions and the scope of the calculations.

Open resource ↗

CATALOGUE

Language models

Documented architecture, parameters and context.

Open resource ↗

PLATFORMS

Compatible hardware

Documented components and configuration rules.

Open resource ↗

RESOURCES

CPU, storage and networking

What the indicators represent and how to calibrate them.

Open resource ↗

Guest Perspective

SOURCES AND DOCUMENTATION

References

EU AI Act — Regulation (EU) 2024/1689 ↗
BOE / DOUE · Original in Spanish
Statute of the Spanish Agency for the Supervision of Artificial Intelligence — Royal Decree 729/2023 ↗
BOE · Original in Spanish
Personal Data Protection and Guarantee of Digital Rights — Organic Law 3/2018 ↗
BOE · Original in Spanish
Spanish National Security Framework — Royal Decree 311/2022 ↗
BOE · Original in Spanish
GDPR compliance of processing that incorporates Artificial Intelligence ↗
AEPD · PDF · Original in Spanish
Guide to audit requirements for processing that includes Artificial Intelligence ↗
AEPD · Original in Spanish
On-Premise vs Cloud: Generative AI Total Cost of Ownership (2026 Edition) ↗
Lenovo Press

Knowledge youcan verify.

Judgement for your infrastructure.

Beyond silicon: judgement and method for bringing AI to your infrastructure

From model to deployment.

How to use the Calculator

How much VRAM does an LLM need? A calculation guide

LLM quantization: FP16 vs INT8 vs INT4

What is the KV cache in an LLM, and why does it consume so much memory?

On-premises vs cloud AI for businesses: when each makes sense

Single server vs AI cluster: when to scale out

AVAILABLE NOW

LLM Calculator

Memory and performance

Language models

Compatible hardware

CPU, storage and networking

Guest Perspective

References

Knowledge you
can verify.