Lab data

Lab live · 5 servers measured · Nov 2026

We run it before we recommend it.

Which server do you actually need for n8n, a local LLM or Supabase? We rent the servers, install the apps, measure RAM, tokens per second and disk speed, and publish the raw numbers with prices after the promo ends.

21.2tokens/s for Llama 3.2 3B on a CPU-only CX43, no GPU

380 MBRAM n8n uses idle; 785 MB right after start

5servers benchmarked so far, raw data public

From the Lab

Measured on servers we paid for.

Browse

Pick what you want to self-host.

How the Lab works

Same scripts. Same day. Every size.

  1. We pay for the servers. Same OS, same scripts, same day for every size.
  2. We install the real app. n8n with Postgres, Ollama with real models, the official Supabase stack.
  3. We publish the raw data. Every table links to the JSON it came from.
  4. We show renewal prices. Not just the promo.