Tether AI Research has introduced a new family of open-source artificial intelligence translation models capable of operating entirely offline on smartphones and laptops. Announced on September 2, the tools target 19 African languages and nine European languages, allowing users to translate text without an internet connection and keep submitted text on their device.
The flagship TranslatePsy-AfriSLM covers 19 African languages, including Hausa, Swahili, Yoruba, Amharic, Zulu, Somali, and Wolof. Companion models AfriNano supports eight African languages, while EuroNano covers nine European languages. Tether said AfriSLM uses around 800 million parameters and in testing outperformed several larger translation models. The company also developed a filtering method that removes up to 96% of low-quality training data.
The release is part of Tether’s broader QVAC strategy to move beyond its USDT stablecoin business into local AI and payment tools, potentially serving communities poorly covered by cloud-based AI services. Tether stated the models could support education and health applications, although additional details were not fully disclosed.
