Tether releases open-source AI translation model, covering 19 African languages and 9 European languages On September 2nd, according to TechFlow, Tether AI Research announced the launch of the QVAC TranslatiPsy series of open-source AI translation models on September 2nd. The models support 19 African languages (TranslatiPsy AfriSLM), 8 African languages (TranslatiPsy AfriNano), and 9 European languages (TranslatiPsy EuroNano). The models can run offline locally on terminal devices such as smartphones and laptops, without the need for internet connection or data upload to third-party cloud servers. Among them, the smallest version of AfriSLM model with only 800 million parameters performed better than larger scale models such as Qwen3.5-122B-A10B and TranslatieGemma-27B in FLORES-200, BOUQuET, and SMOL translation benchmark tests; The minimum version of EuroNano only requires 36MB of storage space, which is about 94% less than Firefox's offline translation configuration. This series of models is currently available for download on Hugging Face, and related research has been included in the EMNLP 2026 conference.