How we test
Last updated
Our method is simple: use the product the way a reader would, measure what matters, and write down the date.
Prices
- Read from each provider's own US pricing page, in US dollars, before tax.
- Every product shows the date we last checked it. Data older than 90 days is flagged for a recheck.
- For hosting, we show both the introductory price and the renewal price when they differ.
Hosting and VPS tests
Each test server runs the same workloads: a self-hosted AI agent, an n8n instance and a small website. We record uptime and response time on a schedule from US and EU locations, and publish the results. [Test setup details and the results dashboard will be added once the test servers are running.]
GPU cloud tests
We rent the same GPU types across providers and measure time to a working instance, model download speed, inference throughput (tokens per second) for a fixed open model, and the final bill, including storage and network charges. [Benchmark details to be added.]
Ratings
We give a score out of 5 only after hands-on testing. Products we haven't tested show "Not rated yet". The score weighs price and value, performance in our tests, reliability, ease of setup, and support and documentation.
Corrections
Prices and plans change often. If you spot something out of date, tell us and we'll recheck it.