UAE’s Technology Innovation Institute launches Falcon 2 rivaling Meta’s Llama 3

The Technology Innovation Institute (TII) of the United Arab Emirates introduces Falcon 2, a new ateration of its large language model (LLM) series with Vision-to-Language Model (VLM) capabilities, along with open-source accessibility, rivalling Meta’s Llama 3 and Google’s Gemma.

Robot hand and laptop

The Technology Innovation Institute (TII) of the United Arab Emirates, one of the leading global scientific research centre and the applied research pillar of Abu Dhabi’s Advanced Technology Research Council (ATRC), has launched the second iteration of its large language model (LLM) the Falcon series, Falcon 2. Within the series, it has unveiled two versions: 

Falcon 2 11B, an efficient and accessible LLM trained on 11 billion parameters with 5.5 trillion tokens and Falcon 2 11B vision-to-language model (VLM), enabling the conversion of visual inputs into textual outputs. Notably, Falcon 2 11B VLM is the first multimodal model in the top-tier market, offering image-to-text conversion capabilities.

The newly launched models have been tested against leading AI models, where Falcon 2 11B has surpassed Meta’s newly launched Llama 3 and rivalled Google’s Gemma 7B, according to independent verification by Hugging Face, a US-based platform hosting an objective evaluation tool for open LLMs. 

TII has made both Falcon 2 11B models open source, aligning with a broader trend in the AI community, giving developers worldwide unrestricted access and contributing to accelerating advancements in AI research and development.

TII aims to expand the Falcon 2 series with models of varying sizes and enhance them with advanced machine learning capabilities like ‘Mixture of Experts’ (MoE). Advanced Technology Research Council Secretary General Faisal Al Bannai, who is also an adviser to the president on strategic research and advanced technology, said they were already working on “Falcon 3 generation”.

The multilingual capabilities of Falcon 2 11B models enable them to tackle tasks in multiple languages, enhancing their effectiveness across diverse industries such as healthcare, finance, e-commerce, education, and legal sectors. Moreover, their ability to run efficiently on minimal hardware makes them highly scalable and adaptable to different infrastructures.

The UAE’s focus on AI has not gone unnoticed, attracting attention from global players and policymakers alike. Emirati AI firm G42‘s strategic shift away from Chinese technology and subsequent $1.5 billion investment from Microsoft.

While the UAE’s AI ambitions have drawn scrutiny, particularly from US officials advocating for allegiance to American or Chinese technology, the country’s strategic partnerships and investments demonstrate a nuanced approach to international relations in the tech sphere.