Learn local AI, no hype
69 guides measured on our machines, organised by topic. Search or filter.
Deployment The VORLUX AI Stack: Every Tool We Use, Nothing Hidden
The open-source tools behind VORLUX AI: Ollama, vLLM, LiteLLM, FastAPI, SQLite, Qdrant, Astro and n8n on two DGX Sparks, with the models we actually run.
No Hype the week in AI, with opinions
Hardware What to buy at each budget, measured.
Models What runs on that hardware, and how well.
Measure and decide Benchmarks, costs and decisions with data.
Measure and decide
Measure and decide
Measure and decide Deployment Getting it running and keeping it running.
Creators Image, video and pipelines on your machine.
Safe and legal EU AI Act, GDPR and grants, as checklists you can complete.
Safe and legal
Safe and legal
Safe and legal
Safe and legal -
The VORLUX AI Stack: Every Tool We Use, Nothing Hidden The open-source tools behind VORLUX AI: Ollama, vLLM, LiteLLM, FastAPI, SQLite, Qdrant, Astro and n8n on two DGX Sparks, with the models we actually run.
Deployment -
Valencia AI Ecosystem: 1,689 Startups and Where Local AI Fits Valencia counted 1,689 startups in September 2025, 28.75% of them in AI. The companies, the public funding and where local AI fits for regional SMEs.
Hardware -
Spain's AI market in 2026: the real figures and where to start How many Spanish companies use AI (INE), what the market forecasts say and who publishes them, which funding is open, and where SMEs should focus.
Measure and decide -
Local RAG over Company Docs with n8n and Ollama Step-by-step tutorial to build a RAG pipeline with n8n, Ollama and ChromaDB on your own hardware, with commands, measured timings and zero cost per query.
Models -
n8n + MCP: Connect Your AI Agents to Any Business Tool MCP passed 97 million monthly SDK downloads in December 2025 and n8n supports it natively. Build AI agents that query databases and read your files.
Deployment -
Mistral Small 24B: Europe's Multilingual, Open-Source Model An honest review of Mistral Small 24B Instruct 2501: model-card benchmarks against GPT-4o-mini, memory needs and where it fits for European SMEs.
Models -
Llama 4 Scout and Maverick: A Practical Review for Local AI Deployment Meta's Llama 4 Scout (10M context, 109B total) and Maverick (1M context, 400B total) bring MoE and native multimodality to open weights.
Models -
Edge AI Hardware Guide 2026: Jetson vs Mac Mini vs NUC Jetson Orin Nano Super, Mac mini M4 and a Core Ultra mini PC compared on vendor specs: memory, TOPS, power and which local models each one can hold.
Hardware -
GDPR Article 25: Why Local AI Inference IS Privacy by Design GDPR Article 25 requires data protection by design and by default. Running AI models on your own hardware is the most direct way to show you meet it.
Safe and legal -
EU AI Act Article 43 Conformity Assessment: an SME Checklist Pick your Article 43 route, open the 13-row document register and run the three Annex VI checks, on the dates the Digital Omnibus set, with free official tools.
Safe and legal -
EU AI Act Compliance Guide 2026: What Spanish SMEs Must Do Now Updated EU AI Act compliance timeline, risk classification, and action plan for Spanish SMEs. Learn how local edge AI helps meet data sovereignty and GDPR.
Safe and legal -
Edge AI in Manufacturing: How Spanish Factories Deploy It Predictive maintenance, visual quality control and operator support with AI on the factory floor: use cases, a starter hardware kit and how a Spanish SME starts.
Hardware -
DeepSeek V3: a 671B Open-Weight Giant for Code and Math DeepSeek V3 packs 671B parameters into a Mixture-of-Experts model that activates 37B per token. In DeepSeek's benchmarks it beats GPT-4o at code and math.
Models -
Best local LLMs: which one to use for your hardware and task Which Ollama model to run for chat, coding, documents, reasoning, vision, Spanish and speech on 8, 16, 32 or 64 GB, with speeds measured on our machine.
Models -
Your First 3 AI Agents: A Local Deployment Guide for SMEs (2026) Deploy three AI agents on hardware you own with Ollama and MCP, in order of safety and impact, starting with a read-only digest on an EUR 80 board.
Deployment -
AESIA: What Every Spanish Business Deploying AI Must Know in 2026 What AESIA is, what it can ask you for, the real EU AI Act dates after the July 2026 Omnibus, and an eight-step checklist to get your AI in order.
Safe and legal -
No Hype: Watermarks, Weights, and Walled Gardens Mistral's ML4 gets a date, OpenAI names its EU watermark, Spanish administrations stall after the pilot, and big tech caps AI spending.
No Hype -
No Hype: Beam, Watermarking, and Local Optimisation Reflection AI's Beam, OpenAI watermarking for the EU, a 125B model on a 12 GB GPU and Google trimming Gemini plans, read with a straight face.
No Hype -
Local AI: what it is, what you need and when it pays off What running AI on your own hardware means, the three pieces you need, where it falls short today and when it costs less than a cloud API.Measure and decide
-
EU AI Act Article 14: Human Oversight for Your Business Complete guide to Article 14 human oversight requirements for high-risk AI systems. 4 core capabilities, implementation checklist, and practical.
Safe and legal -
How to install Ollama and run your first local LLM Install Ollama on Windows, macOS or Linux, run Llama 3.1 (or a 3B model on 8 GB of RAM) and call its API with curl. Measured on our own machine.
Models -
Welcome to VORLUX AI — AI That Stays Where Your Data Lives Learn how this Valencia-based consultancy helps European SMEs deploy private and affordable AI directly on their own hardware to ensure local compliance.
Deployment -
VeriFactu, SII and Crea y Crece: What Your Invoicing Must Do Your VeriFactu date is 1 January or 1 July 2027 (verified on the BOE and AEAT, 9 September 2026), a software checklist, and a local model that checks invoices.
Safe and legal -
Local AI ROI Framework: Calculate Your Cloud Savings in 2026 Calculate the ROI of local AI against cloud APIs with September 2026 prices: TCO tables, break-even by volume, and how to present the case to your CFO.
Measure and decide -
How to Become a Kit Digital Digitalising Agent in 2026 The turnover thresholds, the five documents, the mandatory Kit Digital web page and the Red.es filing, with the programme status verified on 9 September 2026.
Safe and legal -
GGUF Quantization: Pick the Right File for Your Machine How to read a GGUF file name, choose between Q4_K_M, Q5 and Q8 for the memory you have, and measure speed, memory and quality yourself with three commands.
Models -
GDPR and AI in 2026: Local Deployment Is the Clean Answer EUR 7.1 billion in GDPR fines since 2018 and 443 breach notifications a day (DLA Piper, 2026). Why local AI removes the cross-border transfer risk.
Safe and legal -
DeepSeek R1: The Best Open-Source Reasoning Model You Can Run Locally DeepSeek R1 scores 97.3% on MATH-500 in DeepSeek's own paper, and its 14B distilled variant runs on a 16 GB Mac Mini M4. Review, benchmarks and setup.
Models -
How to Classify Your AI's Risk Level Under the EU AI Act (Art. 6) Practical guide to classifying AI systems into the 4 risk categories of the EU AI Act. Decision tree, examples, and checklist included.
Safe and legal -
SLM vs LLM: Why Small Models Are Winning Enterprise AI Gartner expects task-specific models to see three times the usage of general-purpose LLMs by 2027. When to use small vs large models in your business.
Models -
Quantization Explained: Run 70B AI Models on Consumer Hardware GGUF, AWQ and GPTQ explained with real memory numbers: pick a quantization level, check what fits in your RAM, and pull the right Ollama tag in four steps.
Models -
Microsoft Phi-4: The Tiny Giant That Beats 70B Models at Math In Microsoft's own benchmarks the 14B Phi-4 scores 80.4% on MATH, ahead of GPT-4o's 74.6%. What it needs to run locally and where it falls short.
Models -
AITune: Auto-Tune Local Inference on Your NVIDIA Card What NVIDIA AITune really does, when to use it, and the measured Ollama and vLLM settings that speed up a local model on a consumer GPU or a Jetson.
Models -
Automate Code Reviews with AI: n8n + Ollama Workflow Tutorial Build an AI-powered code review pipeline using n8n. Receive PR webhooks, analyze diffs with a local LLM via Ollama, and post results to Discord — all.
Models -
llama.cpp on Intel Arc with SYCL: 16 GB of VRAM on a Budget Build llama.cpp with Intel's SYCL backend, run an 8B model on an Arc A770 or B580, and know from third-party numbers what speed to expect before you buy.
Models -
How to Deploy AI Locally in Your Business: Complete 2026 Guide Step-by-step guide to deploying local AI in your company. Hardware selection, cost comparison, GDPR compliance, and EU grants. No cloud required.
Hardware -
Fine-Tune AI Models on Your Own Hardware: The LoRA Guide for SMEs You don't need a data center to customize AI models. How LoRA and QLoRA let a Mac or a single consumer GPU fine-tune a 7B model, with MLX and Unsloth steps.
Models -
CUDA on Fedora: a Local AI Workstation in an Afternoon Install the NVIDIA driver from RPM Fusion, get Ollama on the GPU in five minutes, then build llama.cpp with the CUDA toolkit using the official Fedora guide.
Models -
AI Evaluations: How to Test Your RAG Pipeline Before Going Live Deploying a RAG system without testing is like shipping code without tests. n8n's AI Evaluations feature lets you measure retrieval accuracy, answer.
Measure and decide -
VORLUX AI Launch Day: We're Open for Business The founding story of VORLUX AI. Why a Valencian consultant built an AI company that runs entirely on local hardware, why the timing is perfect, and how.
Deployment -
NPU vs GPU: Why Neural Processing Units Are the Future of Edge AI When an NPU is enough for local AI inference and when you still need a GPU: vendor TOPS figures, power draw from spec sheets, and what Ollama actually uses.
Hardware -
Local AI Readiness Checklist: Is Your Business Ready? Use this ten-point self-assessment to determine if your company has the necessary infrastructure and data security protocols for local AI deployment.
Measure and decide -
Kit Digital and Spanish AI Grants: 2026 Guide Kit Digital, IVACE and other Spanish public funding for AI projects: what each paid, which calls are closed, and how to prepare for the next window.
Safe and legal -
Google Gemma 4: The Open Model Family That Changed Our Entire Stack A hands-on review of all four Gemma 4 variants — E2B, E4B, 26B MoE, and 31B Dense — with specs, Ollama commands, and honest limits.
Models -
The EU AI Act Calendar After the Digital Omnibus A dated checklist of every EU AI Act obligation after Regulation (EU) 2026/1744, a five-question triage to find your row, and what to finish this month.
Safe and legal -
Score Your Own AI System Against the EU AI Act in 45 Minutes A timed self-assessment: classify the tier in ten minutes, score 18 weighted items in twenty-five, and leave with a number, a band and your three next actions.
Safe and legal -
Article 4 AI Literacy: a One-Page Training Plan A four-session plan you can run this month with free materials and a local model, plus the training record the Commission says is all the evidence you need.
Safe and legal -
ComfyUI Batch Image Generation: Create 100 Product Images in Minutes Tutorial on using ComfyUI's batch generation workflow to create consistent product images, social media graphics, and marketing assets at scale — all.
Creators -
Cloud vs Local AI: Real Cost Analysis for Spanish SMEs in 2026 Compare the expenses of using cloud APIs like OpenAI against running local hardware on a Mac Mini to determine the best AI strategy for your Spanish SME.
Measure and decide -
AI Agents for SMEs: Where the ROI Really Comes From What AI agents do for an SME, three local deployment patterns, why most published agent ROI figures don't hold up, and how to measure your own.
Measure and decide -
The 8 Prohibited AI Practices Under the EU AI Act (With Examples) Article 5 of the EU AI Act bans 8 types of AI. We explain each one with real examples, exceptions, and what to do if you detect one in your business.
Safe and legal -
Spanish AI Grants You Can Apply for in Q4 2026 Which Spanish and Valencian AI funding lines are open in autumn 2026, which wait for a new call, the amounts with sources, and a 30-minute eligibility check.
Safe and legal -
Qwen 2.5 Coder 7B on Your Own Machine in 20 Minutes Hands-on review of Qwen2.5-Coder-7B on Ollama: measured speed and memory, three SME tasks with exact prompts, honest limits, and VS Code and n8n wiring.
Models -
Qwen2.5-72B-Instruct Locally: the 72B vs a 15x Faster 35B What Qwen2.5-72B needs in memory, how fast it really generates on a 128 GB workstation, and a measured Spanish-prompt comparison against qwen3.6:35b.
Models -
A Private Office Voice Assistant with OpenJarvis and Ollama Install OpenJarvis, run a local model via Ollama, add speech in and out with Whisper and Kokoro, and prove with one command that nothing leaves the building.
Models -
Run Your Own AI Agent Gateway with OpenClaw and Ollama in One Command Install OpenClaw, point it at a local Ollama model, schedule your first automation and size the memory it needs, with numbers measured on our own gateway.
Models -
Ollama with MLX on Apple Silicon: What to Measure What Ollama's MLX backend changes on a Mac, which machines qualify, how to tell which runner you are on, and a before-and-after test you can run in ten minutes.
Measure and decide -
Build an AI Content Pipeline with n8n: From RSS to Blog Post Step-by-step tutorial on building an automated content pipeline using n8n. Feeds RSS sources through AI summarization, generates blog posts, and publishes.
Creators -
Build a Self-Improving AI Agent with n8n: The Learning Loop Workflow Learn how to create an n8n workflow that monitors AI agent performance, identifies execution failure patterns, and generates automated improvement steps.
Deployment -
Local AI Hardware Catalogue 2026: What to Buy at Each Budget A buying catalogue for local AI by budget tier, with third-party benchmarks for every device and five questions that tell you which tier you actually need.
Hardware -
Llama 3.3 70B on Your Own Hardware: What It Takes to Run It The memory a 70B model really needs, measured and cited speeds on a 64 GB Mac, two 24 GB GPUs and our GB10, and the SME tasks where it beats an 8B model.
Models -
A 10-Minute Security Checklist for Your Ollama Server Check your Ollama version, update it, bind it to localhost and add an authenticated proxy, with the 2026 GGUF loader CVE as the worked example.
Models -
IVACE Grants for AI Projects in Valencia: What to Prepare INNOVA-CV and DIGITALIZA-CV pay up to 45 percent of an AI project for small Valencian companies; the 2026 windows are closed, so prepare the file now.
Safe and legal -
Google Gemma 3: Open, Multimodal, and It Fits a Mac Mini Learn how to run Google Gemma 3 locally on a Mac Mini to process text and images with privacy and a large context window.
Models -
Where Your AI Data May Live: an EU Data-Residency Checklist Map every AI data flow, pick the legal basis for what leaves the EEA, decide what runs locally, and prove in two commands that your model never phones home.
Safe and legal -
ComfyUI ControlNet Tutorial: Guided Image Generation Step-by-step guide to using ControlNet with Canny edge detection in ComfyUI. Generate images that follow the structure of a reference photo — locally, for free.
Creators -
ComfyUI in 2026: Set Up a Local Product-Image Studio on a 12 GB GPU What the ComfyUI ecosystem looks like in 2026, which image model a small business may use commercially, the hardware it needs and the steps to a local studio.
Creators -
Cloud vs Local AI: Work Out Your Own Break-Even in 15 Minutes Current API prices, measured local speeds and a 12-line script you can run today to see whether a EUR 250 board or a Mac mini pays for itself against the cloud.
Measure and decide -
Point Claude Code at a Local Model with Ollama Three environment variables put Claude Code on a local Ollama model: the setup, what stays on your machine, what does not, and measurements with three models.
Models
Nothing matches. Try another word or remove a filter.
Veredicto semanal
Get new guides before anyone else
Subscribe and we tell you when new guides, templates and workflows go up. One email a week, no spam.
Already published: 69 guides and 25 templates. All free, no signup.
Bonus: the EU AI Act checklist, ready to complete
Once a week No spam Unsubscribe anytime