Foundational LLM · African Languages
Inzwi
Voice for Africa's Languages.
inzwi — Shona for “voice”
Inzwi is a foundational large language model being built natively on Shona, Ndebele, and other African languages. A public API for any developer building multilingual applications — with African languages as first-class citizens, not translations.
Join the Waitlist
Be first for API access. Researchers and enterprise clients both welcome.
Languages in Scope
Shona
Corpus collection & cleaning underway
Ndebele
In scope — data collection starting
Swahili
Planned — Q3 2026
Zulu
Planned — Q3 2026
Yoruba
Planned — Q4 2026
French
ATS school curriculum — Q4 2026
Chinese (Mandarin)
ATS school curriculum — Q4 2026
Spanish
ATS school curriculum — Q4 2026
The Problem We're Solving
Africa's Languages Have Been Left Behind
2,000+
Languages spoken across Africa
<1%
Represented in major LLM training data
16M+
Shona speakers — with almost no AI support
Existing large language models — GPT-4, Claude, Gemini — were trained overwhelmingly on English and European language data. When they attempt Shona or Ndebele, they hallucinate, mistranslate, and misrepresent cultural context.
Inzwi is being built from the ground up with African languages as the primary training target — not a fine-tuned afterthought. Authentic corpora. Native speaker evaluation. Infrastructure that actually works for African developers and their users.
What It Will Do
API Capabilities
Multilingual by Design
Trained natively on African languages — not translated from English. Shona and Ndebele are first-class citizens, not afterthoughts.
Public REST API
A clean, well-documented REST API that any developer can integrate into multilingual web, mobile, or enterprise applications.
African Language NLP
Text generation, translation, summarisation, and Q&A — all with genuine understanding of African linguistic structure and context.
Low-Latency Inference
Optimised for production workloads — fast inference endpoints suitable for real-time applications.
Contextually Accurate
Trained on authentic African-language corpora, not machine-translated text — reducing hallucination and cultural misrepresentation.
Enterprise & Research Tiers
Flexible access tiers for individual researchers, academic institutions, and enterprise clients with SLA guarantees.
Build Roadmap
Where We Are
Data Collection
IN PROGRESSScraping and curating Shona corpora from Wikipedia, news sources, literature, and web sources. Cleaning and deduplication pipeline live.
Corpus Expansion
Expanding to Ndebele, Swahili, Zulu, Yoruba. Synthetic data generation for low-resource language augmentation.
Pre-training
Training the base model on the full multilingual corpus. Architecture optimised for African phonological and morphological patterns.
Fine-tuning & RLHF
Instruction fine-tuning and human feedback alignment. Native Shona and Ndebele speakers involved in evaluation.
API Beta
Private API beta for research partners and early enterprise clients. Documentation, SDKs, and rate-limited access.
Public Launch
Public API with enterprise and research pricing tiers. Full documentation and integration guides.
Who It's For
What You Can Build with Inzwi
Healthcare Communication
Patient-facing health applications that communicate in Shona or Ndebele — critical for rural healthcare access in Zimbabwe.
Education Technology
Learning platforms that teach and assess students in their mother tongue — aligned with Zimbabwe's ATS school curriculum.
Government Services
Digital government services and chatbots that serve citizens in indigenous languages — not just English.
Media & Publishing
Automated translation, summarisation, and content generation for African-language media outlets.
Enterprise Multilingual Apps
CRM, ERP, and customer service tools that operate across English, Shona, Ndebele, and other African languages.
NLP Research
A research-grade model for academics studying African language processing, linguistics, and computational language preservation.
Access Tiers
Built for Both Research and Enterprise
Pricing will be confirmed at API launch. Join the waitlist to be first and lock in early-access rates.
Research Tier
For academic researchers, linguists, and NLP teams studying African languages.
- ✓Generous free tier for non-commercial research
- ✓Access to model weights for academic use
- ✓Co-authorship opportunities on published research
- ✓Direct access to the Inzwi research team
- ✓Priority access to new language releases
Enterprise Tier
For businesses building multilingual applications, products, or services.
- ✓High-volume API access with SLA guarantees
- ✓Custom fine-tuning on your domain data
- ✓Dedicated support and integration assistance
- ✓On-premise deployment option
- ✓White-label licensing available
Help us give Africa its voice.
Whether you're a researcher, a developer, or an enterprise — join the Inzwi waitlist and be part of building the foundational AI layer for Africa's languages.
Or email us directly at support@amaryllissuccess.co.zw