Tether launches open-source offline AI translation models

Tether, the issuer of the USDT stablecoin, has released a suite of open-source AI translation models designed to operate offline on local devices such as smartphones and laptops. The new models—TranslatePsy-AfriSLM, TranslatePsy-AfriNano, and TranslatePsy-EuroNano—aim to improve access to education, healthcare, and critical information in regions with limited internet connectivity.

The TranslatePsy-AfriSLM model supports 19 African languages, including Swahili, Hausa, and Zulu, covering areas that represent nearly half of Africa’s population. Additionally, the TranslatePsy-EuroNano model supports nine European languages and is optimized for multilingual translation with minimal storage requirements.

By functioning locally without reliance on cloud infrastructure, these tools prioritize data privacy and accessibility. The company noted that the smallest TranslatePsy-AfriSLM model outperformed larger frameworks, such as Meta’s NLLB-3. 3B, on specific benchmarks like FLORES-200, despite using only 800 million parameters.