Small, task-specific
open models
The future of open AI isn't only bigger models; it's specialists. We build the full stack, from tokenizers to LLMs to vision, and release it all openly, with African and Nigerian languages as first-class citizens.
Olaverse SDK
The unified Python package for our whole stack: African-language NLP, language identification, diacritic restoration, and the MIST model family, in one import.
Latest Releases
Prismprism-denoiser
Image denoising model that cleans noisy photos while preserving fine detail.
OTK Tokenizersotk-bpe
Byte-level BPE tokenizer built for multilingual text, efficient subword segmentation where general vocabularies fall apart.
MIST Genmist-tg-0.3b
Title-generation model that names chats and documents from their content. Small, fast, and ByT5-based.
DiacNetdiacnet-1.0
African-language diacritic and tone restoration powered by a multilingual ByT5 model, making text readable, searchable, and machine-usable.
31
Open Models
9
Model Families
4,637+
All-Time Downloads
25+
Languages Covered
Model Families
From frontier-merged LLMs to single-purpose specialists, every model belongs to a family with one job.
MIST
Frontier-merged LLMs: from MIST-Mini-8B to MIST-1-140B, including reasoning-tuned Thinking variants.
LID
Language identification: from 5 Nigerian languages to 25 languages worldwide.
DiacNet
Diacritic & tone restoration for YorĂčbĂĄ, Igbo, Hausa and beyond.
Prism
Vision models: super-resolution upscaling, denoising, and image steganography.
MIST Embed
Encoders & sentence embeddings: the retrieval foundation for Nigerian languages.
MIST Rerank
Cross-encoder rerankers, 22.7M to 150M: precision for search & RAG pipelines.
MIST Gen
Task generators: question generation in 25+ languages and title generation.
OTK Tokenizers
Byte-level BPE tokenizers optimised for Nigerian & African-language text.
Legal Peace
Fine-tuned legal reasoning and contract-analysis LLM.
African languages,
first-class citizens of AI
Over 2,000 of the world's languages are African, yet they remain nearly invisible in AI. We're changing that, starting with Nigeria, one focused model at a time. Not with a single giant model, but with a complete, open stack.
Tokenize
OTK byte-level BPE vocabularies built for YorĂčbĂĄ, Igbo, Hausa & Pidgin.
Encode & Embed
A Nigerian ModernBERT encoder and cross-lingual sentence embeddings.
Retrieve & Rank
Compact cross-encoder rerankers for search and RAG precision.
Identify
LID models from zero-dependency to neural, 5 to 25 languages.
Restore
DiacNet brings back the tone marks that make text machine-usable.
See
Prism vision models, upscaling, denoising, steganography.
Open weights.
Zero friction.
Every model loads straight from Hugging Face with the tools you already use: Transformers, sentence-transformers, PyTorch. No API keys, no sign-ups, no rate limits. Fine-tune them, quantise them, ship them.
From the Blog
LatestKnow What Language It Is: Introducing the Olaverse LID Collection
Fast, open-source language identification built for African languages. If you've ever built a chatbot, moderation pipeline, or translation feature for Nigerian users, you've hit the same wall we did: before you can process text, you need to know what language it's in, and most off-the-shelf language detectors fall apart the moment they meet Yoruba [...]
Introducing diacnet-1.0: One Model, Ten Languages, and an Honest Look at What It Can (and Can't) Do Yet
Release notes land on Hugging Face first.
Follow olaverse to catch every drop.
Build With Our Models
Every model is open-weight and free forever. Fine-tune it, deploy it, or drop it straight into your product. If you build something for African languages, tell us about it.