Falcon-Emirati

Falcon-Emirati-7B is a 7B-parameter language model that specializes Falcon-H1-Arabic in Emirati Arabic, the Gulf dialect spoken in the UAE, covering its vocabulary, grammar and cultural context. It is made by the Technology Innovation Institute (TII). The model uses the Falcon-H1 hybrid architecture, which combines State Space Models (Mamba) and Transformer attention in each block. Its training used three sources: authentic Emirati-dialect web data, MSA data about Emirati culture and identity, a