Short answer: the n8n documentation does not publish a fixed minimum RAM or CPU requirement. It explains what memory use depends on and shows benchmarks. This article summarizes what is official (documentation checked on October 15, 2026).
What memory use depends on
According to the Fix memory issues page:
- the amount of JSON data;
- the size of binary data (files);
- the number of nodes in the workflow;
- heavy nodes: the Code node and the older Function node raise memory use a lot;
- manual executions, because n8n copies the data to the interface;
- other workflows running at the same time.
Signs of running out of memory: messages such as “n8n may have run out of memory while executing it”, Connection Lost, 503 Service Temporarily Unavailable or JavaScript heap out of memory in the log. In the Docker image, n8n restarts by itself in those cases. If you are choosing between your own server and the hosted version, see n8n Cloud vs self-hosting.
How to reduce it: process in smaller batches (the documentation gives the example of 200 rows per execution instead of 10,000), avoid the Code node where possible, avoid manual executions with a lot of data and split work into sub-workflows that return little data. You can also raise the V8 memory limit with NODE_OPTIONS=--max-old-space-size=2048 (the value is in MB).
Official benchmarks
From the Measure performance page:
- n8n states it supports up to 220 workflow executions per second on a single instance, and more with multiple instances.
- Single-instance test: an ECS c5a.large instance with 4 GB of RAM, n8n in main mode with PostgreSQL, a simple Webhook + Edit Fields workflow.
- Multi-instance test: seven ECS c5a.4xlarge instances (8 GB each), with 2 webhook instances, 4 workers, a MySQL database and Redis, in queue mode.
These are minimal workflows, so real workflows with AI, files or the Code node use more.
Database and queue mode
- The default is SQLite. PostgreSQL is supported on versions 16, 17 and 18 (docs).
- Queue mode (main + Redis + workers) is the most scalable mode. The documentation does not recommend queue mode with SQLite and says queue mode does not support binary data on the filesystem; for that, use S3 (docs). The full setup is in n8n in queue mode with Redis.
So which VPS?
Since there is no official number, use the calculator for an estimate based on your usage, with the benchmark above (4 GB) as a reference. Once you have a size, compare VPS prices and follow the installation guide with Docker Compose. After it is running, watch real usage (for example with docker stats) and adjust.
Frequently asked questions
What is the minimum RAM for n8n?
The official documentation does not publish a fixed minimum. Memory use depends on data volume, binary files, the Code node and simultaneous executions. The official single-instance benchmark used a machine with 4 GB of RAM.
How many executions per second can n8n handle?
According to the documentation, up to 220 workflow executions per second on a single instance, in a test with a minimal workflow, and more with multiple instances in queue mode.
Can I use queue mode with SQLite?
The documentation does not recommend it. For queue mode, use PostgreSQL and Redis.