The UAE’s Technology Innovation Institute (TII), part of Abu Dhabi’s Advanced Technology Research Council (ATRC), has unveiled two significant advancements in artificial intelligence: Falcon Arabic and Falcon-H1. The announcement was made during the “Make it in the Emirates” event in Abu Dhabi by H.E. Faisal Al Bannai, Advisor to the UAE President and Secretary General of ATRC. Falcon Arabic: A Milestone in Regional AI Leadership Falcon Arabic, the first Arabic language model in the Falcon series, marks a pivotal moment in the advancement of AI in the Arab world. Built on Falcon 3-7B (7-billion-parameter), Falcon Arabic is trained on a high-quality, native Arabic dataset that includes both Modern Standard Arabic and regional dialects. This robust training allows the model to capture the linguistic diversity of the Arab region effectively. According to the Open Arabic LLM Leaderboard benchmarks, Falcon Arabic outperforms all other regionally available Arabic models, matching the performance of models up to 10 times larger. This achievement underscores TII’s leadership in developing sovereign, multilingual AI solutions and highlights the institute's commitment to enhancing local capabilities. Falcon-H1: Redefining Performance and Portability Falcon-H1 is a new AI model designed to address the growing global demand for efficient, flexible, and user-friendly AI systems. It dramatically reduces the computational power and technical expertise needed to run advanced models, thereby expanding access to high-performance AI. Falcon-H1 builds on the success of the Falcon 3 series, which was among the top global AI models capable of operating on a single graphics processing unit (GPU), a crucial breakthrough for developers, startups, and institutions lacking high-end infrastructure. The Falcon-H1 family includes models of various sizes: 34B, 7B, 3B, 1.5B, 1.5B-deep, and 500M. Each model offers a balance between performance and efficiency, allowing users to select the most suitable option for their specific needs. For instance, the flagship 34B model excels in complex tasks, outperforming comparable models from Meta’s LlaMA and Alibaba’s Qwen. Meanwhile, the smaller models can be deployed on edge devices with limited resources, emphasizing the model’s adaptability and inclusivity. Architectural Innovation and Real-World Utility Falcon-H1’s hybrid architecture, combining the strengths of Transformers and Mamba, enables faster inference speeds and lower memory consumption without sacrificing performance. Dr. Najwa Aaraj, CEO of TII, emphasized that this model is not just a research achievement but also an engineering feat, ensuring real-world applicability. The multilingual tokenizer used in Falcon-H1 supports over 100 languages, making it one of the most versatile AI models available. The Falcon-H1 series excels in multiple areas, including mathematics, reasoning, coding, and long-context understanding. Additionally, it sets a new standard for performance-to-efficiency ratios, outperforming models twice its size. This innovation means that powerful AI can now be deployed on edge devices, where privacy, efficiency, and low latency are paramount. Global Applications and Recognition Falcon models have already found practical applications in diverse domains. For example, in partnership with the Bill & Melinda Gates Foundation, TII’s AI has contributed to the development of AgriLLM, a tool that assists farmers in making informed decisions amidst challenging climate conditions. Over 55 million downloads and widespread recognition attest to the Falcon ecosystem's global impact and its status as the most powerful and consistent family of open AI models from the Middle East. TII’s approach to building Falcon models focuses on creating foundational systems that can be adapted to various industry, research, and public-sector requirements. Unlike many AI initiatives that target narrow consumer use cases, TII’s models are designed to be versatile, resource-efficient, and accessible. This strategy ensures that the benefits of AI can be leveraged even in resource-limited settings where ambitious goals and practical constraints coexist. Open Source Commitment and Ethical Guidelines All Falcon models are open source and available on platforms like Hugging Face and FalconLLM.TII.ae. They are released under the TII Falcon License, an Apache 2.0-based license that promotes responsible and ethical AI development. This openness and ethical framework aim to foster a collaborative environment where researchers, developers, and innovators worldwide can benefit from and contribute to the Falcon models, driving further advancements in AI technology. Industry Insights and Company Profile Industry insiders have lauded the launch of Falcon Arabic and Falcon-H1, highlighting their potential to transform the AI landscape in the Middle East and beyond. Dr. Hakim Hacid, Chief Researcher at the AI and Digital Science Research Center at TII, noted that the Falcon-H1 series showcases the innovative potential of ultra-compact models and paves the way for more efficient AI training and deployment. The Technology Innovation Institute (TII) is the applied research pillar of the Advanced Technology Research Council (ATRC) in Abu Dhabi. Established to drive scientific discovery and technological innovation, TII is dedicated to developing cutting-edge technologies that address global challenges. The institute’s focus on creating accessible, high-performing AI models reflects its mission to democratize advanced AI and empower a wide range of stakeholders.