Tether

QVAC TranslatePsy-AfriSLM is an open-source family of small language models built for machine translation across 19 Sub-Saharan African languages , released by Tether's AI research group as part of its QVAC platform. The lineup ships in 800-million, 2-billion, and 4-billion-parameter versions and covers Hausa, Amharic, Yoruba, Lingala, Swahili, Igbo, Zulu, Somali, Oromo, Malagasy, Kinyarwanda, Xhosa, Afrikaans, Wolof, Luganda, Nyanja, Shona, Tswana and Southern Sotho , languages that together account for a large share of the continent's population but are largely underserved by mainstream AI systems. The models are designed to run locally on laptops and smartphones, including offline, and the launch video positions them as tools for classrooms, farmers learning new techniques, and parents accessing educational material in their own language, with a planned pairing with a companion medical model called QVAC MedSci. The accompanying research paper has been submitted to EMNLP 2026 and is available on arXiv , and the release includes roughly 215 million bidirectional training examples paired with English , addressing a well-known scarcity of high-quality parallel data for African languages. For a founder or operator scanning launches, the notable thing here is who is shipping it. Tether, the issuer of USDT and one of the largest holders of US Treasuries in crypto, is now pushing device-local AI through QVAC, a project led by CEO Paolo Ardoino that integrates payments through Tether's Wallet Development Kit, allowing sovereign AI agents to transact in Bitcoin and USDT . This launch extends that thesis from infrastructure into applied models for markets where offline, low-resource inference actually determines whether AI reaches users at all.