Netlify ran one prompt through 11 models and published the spread

Netlify took one prompt and ran it through 11 different large language models, then published the outputs side by side. Same instructions, same input, eleven different results. It hit the front page of Hacker News the same day with over 130 points, which tells you a lot of technical people found it worth stopping for. You can see the full spread in Netlify’s writeup.

Here’s why this matters if you run a business and use AI tools day to day. Most owners pick a model once, based on whatever their tool defaults to or whatever they’ve heard is “best,” and then judge all of AI by that one experience. Netlify’s test shows that’s a mistake. The gap between models on an identical task can be big. Different structure, different tone, different level of detail, sometimes different accuracy. If you’ve ever asked ChatGPT and Claude the same question and gotten oddly different answers, this confirms it’s not random. It’s the model.

The practical takeaway isn’t “switch models.” It’s “test models against your actual work before you decide which one runs it.” A model that’s great for drafting emails might be weak on financial summaries. One that’s strong on code might be clumsy on customer-facing copy. Running the same real prompt through two or three models before committing costs you a few minutes and can save you from building a workflow on a weak foundation.

This kind of side by side comparison is exactly the kind of thing we build into an AI command centre, so a business can see which model actually performs best on their tasks instead of guessing. It’s a simple exercise. Pick your most repeated prompt, run it through a few models, and look at the spread yourself.