About
What WhatRuns is, how it is made, and who pays for it.
WhatRuns is a guide to AI models, hardware and prices, for people who do not already know the words.
Most writing about AI is either marketing or a leaderboard. Neither tells you what will run on the machine you own, what it will cost for the work you actually do, or whether the licence permits you to sell what you build. That is what this is for.
How it works
Everything factual here is dated and read from a source, not recalled.
- Prices are fetched from a public API. They are not typed in, and they cannot quietly drift from the provider's own list.
- Model figures and licences are checked against what each organisation publishes.
- Tool and repository data comes from the GitHub API — stars, licence, and the date of the last commit.
- Anything older than 90 days says so on its own page.
Where a number is arithmetic rather than measurement — memory requirements, for instance — the page says that too, and shows the sum.
What this is not
It is not sponsored. No model, tool or provider pays for placement. Where the free option is the right one, it says so; where you should not bother running anything locally at all, it says that as well.
It is not affiliated with any lab. When a licence forbids commercial use, that goes in the table. When a popular tool turns out to be archived, that goes on the page.
It does not do benchmarks of its own. It tells you how to test models on your own work, which is the only measurement that describes your situation rather than someone else's.
How it is made
Written with AI assistance, and every figure checked against its source before publication. That is worth stating plainly on a site whose entire value is that its numbers can be trusted.
The assistance changes how quickly a draft exists. It changes nothing about the standard: a number without a source does not go in, and a page that cannot be kept current is dated rather than left to rot.