Skip to content
Kalkazu.org
Kalkazu.org

About Kalkazu.org

Kalkazu.org is an independent publication about artificial intelligence, based in Brazil and published in Portuguese and English. We cover language models, autonomous agents and generative AI, with a focus on what changes in practice: what a tool does, how much it costs, where it fails and what that means for people who work with technology.

The monthly Benchmarking

Every month we publish the Kalkazu Benchmarking: the most downloaded and most liked AI models on Hugging Face and the latest releases listed on OpenRouter, with the change from the previous month. Each edition is frozen on the first day of the month and stays available at its own address, with the data open for download as CSV and JSON. If you cite it, the page tells you how.

Who signs the articles

Kalkazu.org is responsible for its articles, not an individual author. That is a stated choice, not an oversight: the site is edited by a small newsroom, and crediting every article to the same person's name would say less about who answers for the content than the brand saying it does.

The articles are presented by Nina, the site's virtual host. She is a character created with artificial intelligence, the same one that appears in our videos, and she is not a person. Articles that carry her name are reviewed by Kalkazu's editorial team.

What takes the place of an author's résumé is our editorial policy: a public document that describes how each article is produced, the role artificial intelligence plays in that process and what happens when we find an error.

What you will find here

  • Analysis of language models: capability, limits, price and use case.
  • Autonomous agents: how they work, where they already help and where they do not yet.
  • Generative AI applied to text, image and code.
  • Articles written for an international audience. The English edition is its own selection, not a mirror of the Portuguese one, and every English article is read by a person before it is published.

How to reach the newsroom

Corrections, story ideas and partnership proposals go to the contact page. Factual corrections are taken seriously and recorded in the article itself; the editorial policy explains how.