Tether has launched TranslatePsy-AfriSLM, a family of free, open-source AI models that translate 19 African languages fully offline, directly on a user's device.
The AfriSLM models use about 800 million parameters, yet each language pair weighs only 21 to 35 MB, small enough to run on a basic phone with no internet connection and no remote server. Tether says that cuts storage requirements by around 94% versus larger translation systems.
They are also fast: roughly 46 milliseconds per sentence, which QVAC claims is up to 78 times quicker than larger models like Salamandra-2B, while matching or beating them on quality.
The 19 supported languages include Swahili, Hausa, Yoruba, Amharic, Zulu, Igbo and Somali, together spoken by hundreds of millions of people. It is part of Tether's QVAC push to put private, on-device AI within reach of users the cloud has largely skipped.
Partner Offers | Get Your Instant BTC Cashback | Your Gateway To Crypto | Arcus Private Whitelist
