The Technology Innovation Institute (TII), a leading global scientific research centre and the applied research pillar of Abu Dhabi’s Advanced Technology Research Council (ATRC), has recently launched Falcon 2, the second iteration of its renowned large language model (LLM).
This series introduces two versions: Falcon 2 11B and Falcon 2 11B VLM. The former is a more efficient and accessible LLM, trained on 5.5 trillion tokens with 11 billion parameters, while the latter, distinguished by its vision-to-language model (VLM) capabilities, enables seamless conversion of visual inputs into textual outputs.
Falcon 2 11B VLM stands out as TII’s first multimodal model and the only one in the top-tier market with image-to-text conversion capability, marking a significant advancement in AI innovation.
H.E. Faisal Al Bannai, Secretary General of ATRC and Strategic Research and Advanced Technology Affairs Advisor to the UAE President, stated in a note, “With the release of Falcon 2 11B, we’ve introduced the first model in the Falcon 2 series. While Falcon 2 11B has demonstrated outstanding performance, we reaffirm our commitment to the open-source movement and the Falcon Foundation. With other multimodal models soon coming to the market in various sizes, we aim to ensure that developers and entities that value their privacy can access one of the best AI models to enable their AI journey.”
Performance and accessibility
Falcon 2 11B has been rigorously tested against several prominent AI models in its class. It surpasses the performance of Meta’s newly launched Llama 3 with 8 billion parameters (8B). It performs on par with Google’s Gemma 7B, achieving near-equal performance scores (Falcon 2 11B: 64.28 vs Gemma 7B: 64.29), as independently verified by Hugging Face, a US-based platform hosting an objective evaluation tool and global leader board for open LLMs.
Notably, Falcon 2 11B and 11B VLM are open-source, empowering developers worldwide with unrestricted access. There are plans to broaden the Falcon 2 next-generation models, introducing a range of sizes further enhanced with advanced machine learning capabilities like ‘Mixture of Experts’ (MoE), aimed at pushing their performance to even more sophisticated levels.
Market impact and multilingual capabilities
All of TII’s AI models released to date have consistently ranked in the top tier globally as the most potent open-source LLMs. The new scaled-down and versatile Falcon 2 11B models are set to achieve greater market adoption in the ever-evolving world of generative AI.
Equipped with multilingual capabilities, Falcon 2 11B seamlessly tackles tasks in English, French, Spanish, German, Portuguese, and other languages. This versatility enriches the models’ effectiveness across diverse scenarios, making them adaptable for global applications.
With its vision-to-language model capabilities, Falcon 2 11B VLM can identify and interpret images and visuals from the environment. It provides various applications across healthcare, finance, e-commerce, education, and legal sectors.
These applications range from document management and digital archiving to context indexing and supporting individuals with visual impairments. Furthermore, these models can run efficiently on just one graphics processing unit (GPU), making them highly scalable and easy to deploy and integrate into lighter infrastructures like laptops and other devices.
Strategic Vision and Future Enhancements
Dr. Hakim Hacid, Executive Director and Acting Chief Researcher of the AI Cross-Centre Unit at TII, added, “AI is continually evolving, and developers are recognising the myriad benefits of smaller, more efficient models. In addition to reducing computing power requirements and meeting sustainability criteria, these models offer enhanced flexibility, seamlessly integrating into edge AI infrastructure, the next emerging megatrend. Furthermore, the vision-to-language capabilities of Falcon 2 open new horizons for accessibility in AI, empowering users with transformative image-to-text interactions.”
The versatility of Falcon 2 11B has led TII to explore more exciting GenAI innovations, including adopting a new type of machine learning capability known as a ‘Mixture of Experts’ (MoE). This method involves amalgamating smaller networks with distinct specialisations, ensuring that the most knowledgeable domains collaborate to deliver highly sophisticated and customised responses. This approach improves accuracy and accelerates decision-making, paving the way for more intelligent and efficient AI systems.
Ethical Considerations and Licensing
Falcon 2 11B is licensed under the TII Falcon License 2.0, a permissive Apache 2.0-based software license that includes an acceptable use policy promoting the responsible use of AI. This ensures that the deployment of Falcon 2 11B and its variants aligns with ethical guidelines and fosters a culture of responsible AI development.
TII’s Falcon 2 series represents a significant leap forward in developing large language models. With its superior performance, multilingual and multimodal capabilities, and commitment to open-source accessibility, Falcon 2 is poised to revolutionise the AI landscape.
The future enhancements, including the Mixture of Experts approach, promise even greater advancements, making Falcon 2 an indispensable tool for developers and organisations worldwide.
