The BaseResearcharXiv
Inherited Weights: Measuring the Tokenization Premium and the Sovereignty Cost of Language Models for the Global South
According to the abstract, Paulo Camara argues that fine-tuned models inherit vocabulary, knowledge, cultural alignment and licence from their base model. The paper measures how much more text costs in tokens across 31 languages and offers an eight-test audit a buyer can run without access to weights.
Why it matters
The same thing, for different readers.
What it means for you
Speakers of smaller languages can pay several times more to use the same AI service, because their words are split into more tokens.
For lawmakers
Governments buying models for local languages get a black-box audit to test how much a model depends on its foreign base model.
For researchers
The paper updates tokenization measurements across 31 languages and 7 tokenizers and shows a Portuguese-built vocabulary lowering the premium for 11 of 13 languages.
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- Tokenization cost measured on 27 articles of the Universal Declaration of Human Rights across 31 languages and 7 tokenizers released in 2023-2024, plus an eight-test black-box audit and an argument about design objectives.
- What it found
- Premiums reach 10x for Amharic in older tokenizers and remain 6.8x in the newest commercial one; Indigenous South American languages pay 1.6x to 2.7x. A Portuguese-built vocabulary lowers the premium for 11 of 13 Latin-script Global South languages.
- What it does not show
- Not stated in the excerpt; the measurements use one parallel text.
- Status
- A preprint: not yet peer reviewed
The record
- Questions
- Who can demand evidence, delay a release or stop an unsafe deployment? · How is AI changing childhood, health and daily life?
- People
- Paulo Camara (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 11, 2026, via
arxiv-ai-safety-query(tier 1: the source’s own channel or an official record) - Cite this
- arXiv, “Inherited Weights: Measuring the Tokenization Premium and the Sovereignty Cost of Language Models for the Global South”, November 10, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/arxiv-2026-11-10-inherited-weights-tokenization-premium