What This Site Is
AI Models Compared is an independent site about the cost and capability of AI coding models and assistants. It exists because the question most developers and teams actually have — which one should I pay for, and what will it really cost me per month? — is surprisingly hard to answer from vendor marketing pages.
Every article here is built on the same two things: figures published by the vendors themselves, quoted exactly and dated, and the arithmetic those vendors do not do for you.
Who Runs It
This is a small, independently operated site. It is not owned by, funded by, or affiliated with OpenAI, Anthropic, GitHub, Microsoft, Google, Cursor, or any other company whose products it covers. No vendor reviews articles before publication, and no vendor has any say in what appears here.
You can reach the person behind the site through the Contact page.
How We Compare Things
Numbers come from primary sources
Benchmark scores, API rates, subscription prices, rate limits and context-window sizes are taken from the vendor’s own documentation, pricing page, or model card — not from other comparison sites, not from press coverage, and not from memory. Where a figure appears in an article, the source it came from is linked.
Published figures are quoted, not paraphrased
When a vendor reports a benchmark result, we reproduce the figure as they published it, including the conditions they attached to it. Where a vendor declines to publish a number, we say that it is unpublished rather than substituting an estimate of our own.
Every figure carries a date
AI pricing and model lineups change faster than almost any other category we could have picked. Articles state the month they reflect, and figures known to come from an older announcement are flagged as such, so you can tell at a glance whether what you are reading is still current.
The arithmetic is ours
Vendors publish a per-token rate or a monthly seat price. What they rarely publish is the break-even point between the two, the real monthly bill at a realistic usage level, or the point at which the cheaper plan stops being cheaper. That calculation is the part we do ourselves, and where the assumptions matter we state them in the article so you can substitute your own.
What We Do Not Do
We do not run private benchmarks and present them as authoritative. Independent evaluation of frontier models is expensive to do properly and easy to do badly, and a comparison site that quietly invents its own scores is worse than useless. Where a benchmark number appears here, it is somebody’s published result and it is attributed to them.
We do not accept payment for favourable coverage, we do not let advertising influence which products are recommended, and we do not publish sponsored articles dressed up as editorial.
How This Site Is Funded
This site carries advertising served by Google AdSense. Ad units are labelled and visually separated from editorial content, and neither Google nor any advertiser has any input into what we write or which products we recommend. We carry no affiliate links and no sponsored content, and have no commercial relationship with any company we write about.
Advertising has no bearing on our conclusions, and no advertiser sees an article before publication. If we ever enter an affiliate relationship, we will disclose it on this page and label the articles it affects.
Corrections
If a figure here is wrong, out of date, or missing the context that makes it meaningful, we would rather hear about it than not. Write to us through the Contact page with the article and the specific number in question, and a link to the primary source if you have one. Substantive corrections are made in the article itself rather than quietly dropped.
Privacy
How this site handles visitor information, including analytics and any future advertising, is set out in full in our Privacy Policy.
Last updated: August 22, 2026.