Making African languages first-class citizens of AI
Olaverse Lab builds small, task-specific open models — with a core focus on African & Nigerian languages. We believe the future of open AI isn't only bigger models — it's specialist models: compact, efficient, and sharply focused on one job.
We started with a question nobody was answering
Why can a language model write a sonnet in French but can't restore the tone marks that make a Yorùbá sentence readable? Hundreds of millions of people speak Nigerian languages every day, yet the AI industry treats them as edge cases.
We decided not to wait for the big labs. Instead of chasing scale, we went the other way: build the complete stack of focused, compact models — a tokenizer that doesn't shred African text, an encoder that understands it, embeddings that can search it, classifiers that can identify it, and restorers that can repair it.
33+ open models later, that stack exists, alongside merged MIST LLMs, rerankers, generators, and Prism vision models. Everything is on Hugging Face, free, forever.
Four principles behind every release
Open by default
Every model we build ships as open weights on Hugging Face. No waitlists, no API keys, no gated demos. If we made it, you can download it.
Small models, sharp focus
The future of open AI isn't only bigger models; it's specialists. One model, one job, done properly: compact, efficient, and cheap to run.
African languages first
Over 2,000 of the world's languages are African, yet they're nearly invisible in AI. We treat Yorùbá, Igbo, Hausa, and Pidgin as first-class citizens, not afterthoughts.
Built in public
We ship openly and often: weights, datasets, demos, and mistakes included. Our track record is public, on Hugging Face, for anyone to verify.
Open research, open weights,
and models built to order
Support the Olaverse ecosystem
The Olaverse is built on open-source principles. Our models stay free because people fund open work — and sponsorship starts with a conversation, not a payment page.
Sponsor a model, dataset, or language
Universities, NGOs, and companies can sponsor a model release, a dataset, or an entire language, with attribution on the work that results. Tell us what you care about and we'll scope it with you.
Build with us
Contribute to our open repositories and join the community.
Let's build together
Whether you want to use our open models, need one built for your business, or share our mission for African-language AI, we'd love to hear from you.