Abu Dhabi takes a decisive step in global artificial intelligence (AI) with Falcon Perception, a multimodal model enabling machines to efficiently see, read and interpret the physical world.
Developed by the Technology Innovation Institute (TII), the applied research arm of the emirate’s Advanced Technology Research Council, Falcon Perception expands the UAE’s AI ecosystem by combining vision and language capabilities.
As global AI competition intensifies, the UAE positions itself among nations capable of advanced multimodal systems at scale, making Falcon Perception central to this ambition. With approximately 600 million parameters, Falcon Perception is notably more compact than many prominent multimodal models, which often use several billion parameters.
“Our goal with Falcon Perception was to challenge the prevailing assumption that vision systems must rely on complex multi-stage architectures. By demonstrating that a single dense transformer can handle perception tasks efficiently, we are opening the door to a new generation of scalable multimodal systems,” said Hakim Hacid, chief researcher at TII’s Artificial Intelligence and Digital Research Centre.
This efficiency-performance balance demonstrates a broader AI trend: rather than increasing parameter counts or requiring extensive compute resources, researchers emphasise model design optimisation, such as efficient transformer variants, to achieve strong results even on resource-constrained hardware.
Multimodal AI is widely seen as the next frontier of artificial intelligence. While large language models (LLMs) have dominated recent advances, the ability for machines to interpret and interact with the physical world is becoming critical as AI expands into robotics, manufacturing and intelligent infrastructure.
Falcon Perception employs a unified transformer-based architecture, enabling end-to-end integration of visual and linguistic features at the model input level. Unlike traditional pipelines that join separately trained computer vision and NLP modules, Falcon Perception processes and reasons across modalities directly within its shared network, reducing inference latency and deployment complexity.
Consequently, the system interprets complex, multi-object visual scenes using natural language prompts. Users can instruct the model to identify, count or segment specific objects within an image, and Falcon Perception returns bounding boxes, segmentation masks or text outputs, even in crowded, intricate environments.
Such capabilities have clear implications for industry. In manufacturing, the model could enable automated inspection and defect detection. In robotics, it enables machines to follow natural-language instructions in dynamic environments. In enterprise settings, it can streamline large-scale document processing and visual data labelling.
For TII, the launch represents not only a technical milestone but also a step in a broader national strategy. Since beginning its AI agenda, the UAE has prioritised building sovereign capabilities, ensuring domestic development, responsible governance and alignment with long-term economic goals for critical technologies.
“From the outset, Abu Dhabi has focused on advancing AI research that is safe, transparent and sovereign by design,” said Ray O Johnson, chief executive of TII. “Our goal is to ensure governments, industry and society can adopt AI with confidence.”
TII’s work spans AI safety, evaluation and deployment frameworks, and large-scale research programmes. A flagship outcome of this effort is Falcon, the UAE’s homegrown LLM, first launched by TII in 2023. Falcon quickly gained international attention for its performance and for being released as an open source model, reflecting Abu Dhabi’s belief that openness and governance can coexist.
Falcon is not positioned merely as a technical achievement, but as part of a broader national AI development system. By combining scientific research with agile decision-making at a government level, Abu Dhabi aims to accelerate adoption while maintaining oversight and trust.
