Guides to AI hosting

A checklist that matches each self-hosted n8n risk to a setting, with n8n's 2026 security advisories and what four hosts' ready-made n8n servers set up.

n8n's own figures on RAM and CPU: memory on each n8n Cloud plan, n8n's load test on a 4 GB server, and which server size to pick for your workflows.

How fast small models run on a processor alone, from LocalScore's public results, what fits in 8, 16 and 24 GB, and how to set up Ollama safely.

n8n Cloud against a self-hosted server for three example workloads at this week's prices, with the model bill and the costs people forget.

Five servers for self-hosted n8n, each matched to a need: memory for the money, one-click installs, hourly billing, dedicated cores and EU data centres.

A plain checklist that matches each OpenClaw risk to a setting, from OpenClaw's own security, update and backup guides, with the hosts' one-click set-ups.

A calculator that adds this week's server prices to a model API at your own usage, with what drives the token bill and when a subscription can pay.

Hermes Agent or OpenClaw: chat apps, security model, update pace and server needs, compared from both projects' own documentation and GitHub data.

From an empty server to n8n with HTTPS, a task runner, daily backups and updates you can undo, using n8n's own Docker and Caddy example, step by step.

Dify, Langflow and n8n on your own server: what each is for, licences, the server each needs, install steps, and what to do now that Flowise has ended.

Host a remote MCP server on a small VPS or with n8n: the transport, step-by-step set-up with the official SDK, authentication and a security checklist.

Mac mini or rented server for OpenClaw: the Mac-only features, the server-plus-Mac set-up for iMessage, electricity from Apple's figures and VPS prices.

We compare like with like: the same open model through an API, on a rented GPU and on a card you buy, with break-even points from current prices.

How much GPU memory popular open models need at 16-bit, 8-bit and 4-bit, the extra that long conversations take, and which rentable GPUs fit each one.

Kimi K3 has 2.8 trillion parameters. We work out the GPUs it and DeepSeek V4 need, the hourly cost of the machines and when an API makes more sense.

Which rented GPU holds Llama 3.3 70B at each precision, the speed each card can reach by NVIDIA's own figures, and the Ollama commands to run it.

gpt-oss-120b needs one 80 GB GPU, gpt-oss-20b a 16 to 24 GB card. The memory each needs, the cards to rent and the Ollama and vLLM commands to run them.

Tokens per second for open models on the H100, H200, A100, L40S, RTX 4090 and other GPUs, from public llama.cpp benchmarks, with sources and dates.

Which Qwen3.8-27B files fit on one 24 GB or 32 GB GPU, how long a conversation each allows, the speed to expect and the commands to run it.

Which serving engine to run on a rented GPU: published tests of vLLM, Ollama and SGLang, what each suits, and the set-up commands for all four.

What server a private chat assistant for 5 to 20 people needs, Open WebUI or AnythingLLM, and step-by-step set-up with logins and HTTPS.

Monthly GPU servers from GPU Mart, Unihost and Liquid Web: cards, data centres, terms and fees, and when a monthly server beats paying by the hour.

How much quality 8-bit, 6-bit, 4-bit and 2-bit files lose, by llama.cpp's own tests, and the settings that make a self-hosted model answer worse.

Runpod, Novita, DigitalOcean, UpCloud and Vast.ai compared from their own pages: billing steps, stopped-machine charges, locations and live prices.

OpenAI is retiring custom GPTs on 11 December 2026. What moves to a ChatGPT plugin, what you lose, and two document chatbot services to use instead.

Fine-tune an open model with Unsloth on one rented GPU: the memory each size needs, the dataset format, a short training script and export to Ollama.

Published LoRA and QLoRA runs took hours on one GPU; a tuned model online runs 730 hours a month. Run times, memory by method and live GPU prices.

CustomGPT.ai, DocsBot or AnythingLLM on your own server: limits, cited sources, refreshing, data storage and training terms, with live prices.

GPU memory needed by Flux, SDXL, Wan, LTX and HunyuanVideo at common settings, from the makers' own files and figures, and which rentable GPUs fit.

Flux LoRA training with AI Toolkit on a rented 24 GB GPU: captions, install, the settings to change and testing the result in ComfyUI.

Minutes per clip for Wan 2.2 on the RTX 4090, A100 and H100 from Wan's own tests, what LTX and HunyuanVideo need, and how to compare cost per clip.

For n8n servers, GPUs, chatbots and model APIs: where 20 providers store your data, where each company is based and what the GDPR asks of you.

Which providers let you cap spending, prepay or get alerts, what still bills when a server is stopped, and how to stop an n8n agent from looping.