Abu Dhabi's TII Debuts Falcon 2 11B, Advancing AI With Open-Source Models
The Technology Innovation Institute (TII), a pivotal entity in global scientific research and the applied research arm of Abu Dhabi's Advanced Technology Research Council (ATRC), recently unveiled the second version of its large language model (LLM), known as Falcon 2 11B. This launch marks a notable advancement in the field of artificial intelligence (AI), positioning the institute at the forefront of AI innovation.
Falcon 2 emerges in two distinct forms: Falcon 2 11B and Falcon 2 11B VLM. The former is an enhanced, efficient LLM trained on a staggering 5.5 trillion tokens and boasts 11 billion parameters. The latter, Falcon 2 11B VLM, introduces an innovative vision-to-language model (VLM) feature, converting visual data into textual descriptions. This model is recognized for its multilingual capabilities and stands as TII's initial foray into multimodal models, as well as the sole model in the top tier market capable of image-to-text conversion.
Independent evaluations by Hugging Face, a US-based platform for open LLMs, place Falcon 2 11B among the elite in its category. It not only surpasses Meta's Llama 3 but also matches the performance of Google's Gemma 7B, demonstrating TII's commitment to leading AI advancements. Further enhancing their appeal, both Falcon 2 11B and its VLM counterpart are open-source, inviting developers globally to engage with these top-tier models without restrictions.
The Institute's AI models, including the versatile Falcon 2 11B, have consistently secured top rankings globally as the most robust open-source LLMs available. Their multilingual functionality caters to a variety of languages, including English, French, Spanish, and German, among others. The Falcon 2 11B VLM model, in particular, offers wide-ranging applications across multiple sectors, from healthcare to education, by efficiently interpreting images and visuals into descriptive text.
His Excellency Faisal Al Bannai, Secretary General of ATRC, and Dr. Hakim Hacid, Executive Director and Acting Chief Researcher of the AI Cross-Centre Unit at TII, both emphasized the significance of these models. They highlighted their sustainability, efficiency, and the opening of new horizons for AI accessibility through their vision-to-language capabilities. The models' scalable nature, requiring only a single GPU for operation, underscores their adaptability and potential for broader application, including in GenAI innovations and the exploration of Mixture of Experts (MoE) machine learning capabilities.
Both models are licensed under the TII Falcon License 2.0, an Apache 2.0-based software license promoting the responsible use of AI. This aligns with TII's broader goal of ensuring privacy and accessibility for developers and entities embarking on their AI journey, further solidifying TII’s position as a leader in the open-source AI movement.
With the introduction of Falcon 2 11B and its subsequent models, TII not only reaffirms its commitment to advancing AI technology but also to fostering an environment where innovation can thrive. This approach aims at developing AI systems that are not only more intelligent and efficient but also accessible and responsible, leading to a future where AI's potential can be fully realized across various industries.
Sign in with Google