Mansa, The Enterprise Engine for African Language AI

Developed by All Lab, Mansa brings deep contextual understanding to 30 + African languages, powering translation, communication, and content generation with enterprise-grade precision.

Powering Global Innovation with African Language AI
African Languages Lab (All Lab) advances African language AI to make technology and markets more accessible.

The large-scale AI model built to understand text across African languages, powering intelligent solutions rooted in African contexts.

Advanced training processes that teach AI to learn from African language data, ensuring accurate, culturally aware, and scalable intelligence.

Every language tells a story. At All Lab, we believe that every African language deserves a voice in the digital world. What began as a research initiative to understand Africa’s linguistic diversity has grown into a foundation powering enterprise-ready AI.

From the bustling streets of Lagos to remote villages across the continent, our work connects people, businesses, and governments to African languages through intelligent technology. We build machine translation systems, large language models like MansaLLM, and data infrastructure that not only understands words but the culture and context behind them.

Our mission is clear: make African languages accessible and usable in technology, reduce entry barriers for enterprises entering African markets, and ensure that African voices shape global innovation. With award-winning models and expanding coverage across 40+ languages, All Lab is both a guardian of heritage and a driver of progress.

All Labs
Powering Global Innovation with African Language AI

https://www.africanlanguageslab.com/
https://all-lab-portal.com/translate-tool

: A small language model for low-resource African languages

As AI practitioners, we are committed to forging an inclusive future through the power of AI. While AI holds the promise of global prosperity, the challenge lies in the resources required for large models, which are often out of reach for the majority of the world and fail for the languages in those contexts. Open-source models have attempted to bridge this gap, but more can be done to make models cost-effective, accessible, and locally relevant. Introducing InkubaLM (Dung Beetle Language Model) – a robust, compact model designed to serve African communities without requiring extensive resources. Like the dung beetle, which moves 250 times its weight, InkubaLM exemplifies the strength of smaller models. Accompanied by two datasets, InkubaLM marks the first of many initiatives to distribute the resource load, ensuring African communities are empowered to access tools such as Machine Translation, Sentiment Analysis, Named Entity Recognition (NER), Parts of Speech Tagging (POS), Question Answering, and Topic Classification for their languages.

Model
To address the need for lightweight African language models, we introduce a small language model, InkubaLM-0.4B, trained for the five African languages: IsiZulu, Yoruba, Hausa, Swahili, and IsiXhosa. During training, we also include English and French.

InkubaLM-0.4B has been trained from scratch using 1.9 billion tokens of data for the five African languages, along with English and French data, totalling 2.4 billion tokens of data. Similar to the model architecture used for MobileLLM, we trained InkubaLM with a parameter size of 0.4 billion and a vocabulary size of 61788. The figure below shows the training data and model sizes of different public models. When we compare our model in terms of these parameters, we find that our model is the smallest in terms of size and has been trained using the smallest amount of data compared to other models.



https://lelapa.ai/inkubalm-a-small-langu...languages/
https://huggingface.co/lelapa/InkubaLM-0.4B

AI and data services built for today’s most capable multimodal systems.

Powerful AI starts with data. Awarri provides end-to-end data services that power today’s most advanced systems, from labelling to training to human feedback.

N-ATLaS-LLM - Multilingual African Language Model
N-ATLaS-LLM is a fine-tuned multilingual language model based on Llama-3 8B, specifically designed to support African languages, including Hausa, Igbo, and Yoruba alongside English. This model is powered by Awarri Technologies an initiative of the Federal Ministry of Communications, Innovation and Digital Economy as part of the Nigerian Languages AI Initiative to promote digital inclusion and preserve African linguistic heritage in the digital age.

Model Overview
N-ATLaS-LLM is built on the Llama architecture and has been fine-tuned on over 400 million tokens of multilingual instruction data. The model demonstrates strong performance across multiple African languages while maintaining excellent English capabilities.

Key Features
Multilingual Support: Native support for English, Hausa, Igbo, and Yoruba
Cultural Relevance: Trained on culturally relevant content from Nigerian sources
Instruction Following: Fine-tuned for instruction-following tasks
Tool Integration: Built-in support for tool integration capabilities

www.awarri.com
https://huggingface.co/NCAIR1/N-ATLaS

Building Agentic AI with the most Inclusive Datasets for Africa
We leverage hyperlocal multimodal datasets to develop powerful language models and AI agents that truly understand and speak African languages.

Building Truly Inclusive Agentic AI
EqualyzAI is one of Africa’s fastest-growing AI startups, dedicated to democratizing artificial intelligence for the continent’s diverse linguistic communities. We specialize in collecting hyperlocal, multimodal datasets to develop powerful domain-specific Small Language Models (SLMs) and inclusive AI agents. Our mission is to unlock opportunities for native speakers by enabling them to interact with technology in their mother tongues.
At EqualyzAI, we are pioneering the development of truly inclusive Agentic AI solutions that can think, reason, understand, and respond in African languages. By leveraging hyperlocal datasets—collected in collaboration with native language speakers—we ensure our AI models are deeply rooted in the cultural and linguistic contexts of the communities they serve. This approach enables us to create AI systems that are not only technologically advanced but also socially and culturally aligned with the diverse populations across Africa.

Our commitment is to unlock AI possibilities for over one billion native dialect speakers by harnessing hyperlocal, multimodal datasets to build agentic AI—powerful small language models and intelligent agents—that truly understand, speak, and uplift Africa’s diverse languages and dialects.

https://equalyz.ai/

Jacaranda launches open source LLM in five African languages

Last week, we expanded UlizaLlama (AskLlama), our open-source Large Language Model (LLM), to provide AI-driven support in multiple African languages, including Swahili, Hausa, Yoruba, Xhosa, and Zulu. The new multi-lingual model will help deepen how we support new and expectant mothers at scale, while carving new in-roads for AI-driven services across Africa in other sectors.

How does a multi-lingual LLM support new and expecting mothers across Africa?
Off-the-shelf Large Language Models, or LLMs, are typically ineffective in low-resource settings, in part because they’re not adapted to work in languages with limited training data, or customized to specific ‘domains’, like health, agriculture, or education.

In October 2023, we developed the world’s first Swahili-speaking LLM to address this challenge. Our technology team extended the capabilities of Meta’s Llama2, trained the model to respond to general Swahili queries, and then customized it to work within our use case – personalized mHealth support for Kenyan mothers.

In July 2024, we extended this model to Hausa, Yoruba, Xhosa, and Zulu, to reflect our scale ambitions for PROMPTS into Nigeria and South Africa, and as a stepping stone towards our broader ambition of reaching all mums with lifesaving information. Our tech team accomplished this by replicating the process used in the Swahili LLM development: pre-training Meta’s Llama3 for each language, merging the pre-trained models, and finetuning the combined model to create multiple Multilingual LLMs.

We saw promising results in medical accuracy, fluency, and contextual coherence – and we have subsequently integrated the model into our digital health platform, PROMPTS.

https://jacarandahealth.org/jacaranda-la...languages/
https://huggingface.co/Jacaranda/UlizaLlama

We introduce CroissantLLM, a 1.3B language model pretrained on a set of 3T English and French tokens, to bring to the research and industrial community a high-performance, fully open-sourced bilingual model that runs swiftly on consumer-grade local hardware. To that end, we pioneer the approach of training an intrinsically bilingual model with a 1:1 English-to-French pretraining data ratio, a custom tokenizer, and bilingual finetuning datasets. We release the training dataset, notably containing a French split with manually curated, high-quality, and varied data sources. To assess performance outside of English, we craft a novel benchmark, FrenchBench, consisting of an array of classification and generation tasks, covering various orthogonal aspects of model performance in the French Language. Additionally, rooted in transparency and to foster further Large Language Model research, we release codebases, and dozens of checkpoints across various model sizes, training data distributions, and training steps, as well as fine-tuned Chat models, and strong translation models. We evaluate our model through the FMTI framework, and validate 81 % of the transparency criteria, far beyond the scores of even most open initiatives. This work enriches the NLP landscape, breaking away from previous English-centric work in order to strengthen our understanding of multilinguality in language models.


https://huggingface.co/croissantllm

Do more with Grok.
Unlock a SuperGrok subscription on Grok.com.

We've just launched SuperGrok Heavy, providing access to Grok Heavy and much higher rate limits.

https://x.ai/
https://grok.com/

Here's an image I generated on Galaxy.ai using Nano Banana Pro. Here's the prompt I used which I made up myself and didn't use the prompt library to create.

Quote:create a picture of a rock band performing live in concert outdoors

Here's the prompt I used with Google's Nano Banano Pro on Galaxy.ai from their prompt library.

Here's the prompt library prompt:

Quote:Turn off the main lights and illuminate the scene with warm candlelight to create a cozy New Years atmosphere. Add festive New Years decorations in red and green throughout the table and home-seasonal greenery, wine bottle, ornaments, ribbons, candles and subtle accents-along with celebratory dishes arranged on the table, keeping the mood elegant, intimate, and inviting.

Self-Attention: This core mechanism lets the model focus on different parts of the input text to understand context, figuring out which words are most relevant to each other, even if they're far apart.
Parallel Processing: Unlike older models that processed words one by one, Transformers process entire sequences at once, drastically speeding up training on massive datasets.
Encoder-Decoder Structure: They typically use encoders to understand input and decoders to generate output, though some LLMs use only decoder-style blocks.
Tokens: Text is broken down into "tokens" (words or sub-words) that are converted into numerical vectors, allowing the model to process language mathematically.
Key Characteristics of LLMs
Massive Scale: LLMs have billions of parameters and are trained on enormous amounts of text and data from the internet, books, and more.
Pre-training & Fine-tuning: They learn general language patterns during broad pre-training and can then be specialized (fine-tuned) for specific tasks.
Generative: They predict the next most likely token, allowing them to generate coherent and creative text, code, or even images and audio.

Watch this video for a visual explanation of the Transformer model:
https://youtu.be/k1ILy23t89E?si=UlwtKorH1rEkhEDM