Legal & corrections
AI Model Directory is an informational project published by appana Technologies, a UAE AI company that builds and funds it, and independent of every company it covers. This page explains what it is, what it isn’t, who is responsible for it, where its data comes from, and how to get something corrected.
This site is in beta
In practice that means a few specific things:
- The directory covers 100 models today. Models are added and retired as the field moves, so a model you looked at last month may not be here now.
- A model is only given an overall score when independent measurements cover at least 80% of its rubric’s weight. Many models, and whole categories such as audio, search, and enterprise tools, are listed without a score rather than given an estimated one.
- Scores move when the underlying leaderboards, prices, and published results are refreshed. A number you cite today can be different next week.
- Sections, wording, and layout are still changing, and links may move.
Treat everything here as a work in progress and verify anything you plan to act on. Corrections are welcome at the address below.
Information only, not advice
Everything on this site is general information to help you research AI models and tools. It isn’t professional, legal, financial, procurement, or investment advice. Test any product against your own needs before you rely on it.
The site is provided “as is”. We work to keep it accurate, but AI products change quickly and we can’t guarantee that every detail is complete, current, or error-free. To the extent the law allows, we’re not liable for decisions made on the basis of this site.
Opinions, scores, and facts
The strengths and weaknesses on each card are editorial opinions based on publicly available information at the time of writing. They’re not claims about any company’s intentions or conduct.
Scores are calculated from third-party data using the rules published in the scoring methodology. Every mark links to its source and shows the date of the data. A score reflects that data and those rules, not a guarantee of how a product will perform for you.
Independence and trademarks
The site is built and funded by appana Technologies, which makes none of the models listed here and has no stake in any of them. Beyond that we’re not affiliated with, sponsored by, or endorsed by any company or product named on this site. No company pays to be listed, scored, or placed, and nobody — appana included — reviews content before it’s published.
Product and company names are trademarks of their respective owners, used only to identify the products being described. The publisher’s logo identifies who designed and developed this project; other logos shown belong to organisations credited as having helped build a listed model. Each is reproduced from the mark its owner publishes, and no maker’s logo is used to rank or endorse anything.
Data sources and licences
We summarise, rank, and convert third-party data into marks. We don’t claim ownership of it, and each source is credited where its data appears.
- LMArena leaderboard data ↗CC BY 4.0
Human-preference ratings behind scores and “top-rated” model chips.
- Anthropic Economic Index ↗Data released under CC BY by Anthropic
Adoption shares, country adoption, and time estimates on the Use cases tab.
- Natural Earth ↗Public domain
Country outlines behind the world map on the Use cases tab.
- Vectara hallucination leaderboard ↗Apache 2.0
Factual consistency rates behind accuracy scores.
- SWE-bench Verified leaderboard ↗Public research results; cited with a link
Published results behind code-quality scores.
- OpenRouter public models API ↗OpenRouter’s documented public API, used under its terms
List prices, context windows, and supported features behind cost, context, and deployability scores.
Research studies and vendor announcements are cited with links wherever their figures appear. Quoted figures remain the work of their authors.
Privacy
You can choose to give us your email address for the weekly Executive Model Brief, delivered on Monday mornings. We use it only to confirm and deliver that briefing through Resend, our email provider. We don’t sell subscriber data. You can unsubscribe from any issue; an address that has not been confirmed is not added to the list. We also use Vercel Web Analytics to count page views and accepted briefing signup requests. Signup analytics contain only the interface language, never your email address or confirmation token; a request count is not a confirmed subscriber count. Vercel states that it works without third-party cookies, identifies visits by a hash of the request that is discarded after 24 hours, and gives us only aggregated statistics, such as pages viewed, referrer, country, browser, and device type. It doesn’t let us identify you. Vercel’s analytics privacy details
The site is hosted by Vercel, which processes standard technical data such as IP addresses to serve pages.
Corrections
If something here is wrong or out of date, we want to fix it. Corrections are made on the strength of evidence, such as a public source, documentation, or published results, and scores are only changed by updating the underlying data or the published method, never by hand.
For corrections, questions about how something was scored, or legal requests, email reachout@appana.ae. Please include the page, what’s wrong, and a link to your evidence.