Skip to content
About

Somali deserves AI built in Somali.

Somast is a research company in Mogadishu. We build Somali language data and models, teach people to work with them, and publish for others to use.

Where we are today

Each figure shows the date it was last checked.

Full company profile

20M+

people speak Somali

Oct 1, 2026

3

open datasets

Oct 1, 2026

1.1B+

corpus tokens

Oct 19, 2026

5

bootcamp cohorts

Oct 1, 2026

200+

bootcamp alumni

Oct 1, 2026

The AI data gap

Somali is ~0.005% of language-identified web data, less than 0.01% of English's share

More than 20 million people speak Somali, yet it barely registers in the datasets that train the world's AI. That gap is why Somast exists.

Source: Common Crawl primary-language page distribution (CLD2, aggregated crawls through CC-MAIN-2026). Percentages are share of language-identified pages. Somali vs English ratio ≈ 0.01% of English's page share.

English
45.5%
Chinese
5.6%
German
5.7%
French
4.6%
Russian
6.9%
Spanish
4.5%
Arabic
0.65%
Swahili
0.008%
Somali
0.005%
Somast++1.1B tokens

Open Somali text we've released, not shown on the same scale as global crawl percentages above.

Our story

Why we started

Somali is spoken by more than 20 million people, yet it holds an almost invisible share of the data that trains modern AI, a fraction of a percent of what English alone contributes. Every model, tokenizer, and benchmark we tried to use for Somali either didn't exist or quietly failed on the language's morphology and dialects.

So in 2023 we started building the missing layer ourselves. We released the first version of Soplang, a programming language with Somali syntax, and began teaching data, programming and machine learning to Somali learners on YouTube. In 2024 we published our first open corpora and started working with partners. In 2025 we released Soplang v2 and launched talent programs so the people who will build Somali AI next could learn the skills to do it themselves.

Since then we have released two versions of Soplang, published our first open corpora, launched talent programs and run five bootcamp cohorts with more than 200 alumni. we have released Somali NLP Corpus 1.0.0 and Soplang 3, and our data and machine learning services are now open. We are still near the start of closing the gap.

What we do

  • Data

    Somali text and speech corpora with documented sources and licences.

    Browse the datasets
  • Models

    A Somali tokenizer, speech-to-text checkpoints and translation previews.

    Try the models
  • Education

    Sponsored bootcamps and mentored learning tracks for Somali engineers and analysts.

    See the bootcamps
  • Technology consulting

    Software and advisory work for partner organisations, including websites and mobile apps.

    About our client work

Organisations can also ask us to build, host and maintain their software. See our services

Research philosophy

Every language needs AI infrastructure. We build it for Somali first, and to a standard the field can trust.

Documented and shareable

Datasets, model details, and evaluation methods are documented so others can verify and build on our releases.

Somali-led

Built with Somali speakers, who shape the technology and its direction.

Rigorous & reproducible

Clear benchmarks, documented methods, and results others can reproduce and extend.

Our approach

From raw data to published models in three steps

01

Collect & annotate

Community contributors, radio archives, and web crawls feed raw Somali text and speech into our pipeline. Every source is documented and licensed.

02

Clean, align & train

Data is deduplicated, filtered for quality, and used to train tokenizers, ASR models, and language models.

03

Publish

Datasets, model checkpoints, evaluation code, and research notes are published under permissive licenses for anyone to use, study, and improve.

Roadmap

How the lab is evolving

2023

Somast founded

Started in Mogadishu to build the datasets, models and tools Somali needs for AI.

2023

Soplang v1 released

First release of Soplang, a programming language with Somali syntax.

2023

Open education started

Began teaching data, programming and machine learning to Somali learners on YouTube.

2024

First open corpora released

Published the first Somali text and speech datasets under open licences.

2024

Partnerships launched

Started working with organisations that use or support Somali language technology.

2025

Soplang v2 released

Second major release of Soplang, the Somali-syntax programming language.

2025

Talent programs launched

Launched bootcamps and fellowships that train local engineers in machine learning and NLP.

2026

Data and machine learning services

Custom data and machine learning work: we prepare data and build models for your needs.

2026

Release Somali NLP Corpus

Full release of our Somali text corpus, with documentation and the licence of each source.

2026

Release Soplang v3

Third major release of Soplang, the Somali-syntax programming language.

2027

Launch mentorship

A mentorship program that connects learners with experienced practitioners.

Sharafdin Yusuf

Sharafdin Yusuf

Lead Engineer & Researcher

Somali people should build, shape, and own the systems that reflect their language and culture, as well as use AI tools made elsewhere.

Documented and shareableSomali-firstReproducibleSomali-ledLow-resource NLP

Ways to take part

Choose the path that fits you.

[email protected] · Mogadishu, Somalia · Last updated Oct 1, 2026