UAE unveils Falcon Perception in push for AI independence

Abu Dhabi’s Technology Innovation Institute (TII) has unveiled Falcon Perception, a compact multimodal transformer model designed to enable machines to see, read and interpret the physical world while advancing the UAE’s ambition for sovereign AI capabilities. Described by TII as a unified vision-and-language system with approximately 600 million parameters, Falcon Perception is intended to reduce inference latency and deployment complexity compared with larger, multi-stage multimodal architectures that often run into 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.

Falcon Perception uses a single, end-to-end transformer-based architecture that integrates visual and linguistic features at model input, rather than joining separately trained computer vision and NLP modules. That approach allows the system to process and reason across modalities within a shared network, enabling users to instruct the model with natural-language prompts and receive bounding boxes, segmentation masks or text outputs for objects in complex, multi-object scenes — including identification, counting and segmentation tasks in crowded environments.

Capabilities and potential uses

• Unified vision-and-language processing in a single dense transformer.
• Outputs include bounding boxes, segmentation masks and textual descriptions prompted by natural language.
• Designed for lower compute and deployment complexity compared with billion-parameter multimodal models.
• Targeted applications include automated inspection and defect detection in manufacturing, instruction-following in robotics, and enterprise-scale document processing and visual data labelling.

TII positions Falcon Perception as part of a broader strategy to build domestic AI capabilities that can be deployed across real-world industries. The institute — the applied research arm of Abu Dhabi’s Advanced Technology Research Council — has emphasised work on AI safety, evaluation and deployment frameworks alongside large-scale research programmes. Falcon, the UAE’s homegrown large language model first launched by TII in 2023 and released as open source, is cited as a flagship outcome of this effort.

“Falcon Perception reflects TII’s commitment to advancing AI capabilities that are both cutting-edge and practical. By rethinking how vision and language models are built, we are enabling more efficient multimodal systems that can be deployed across real-world industries while strengthening sovereign AI capabilities,” said Najwa Aaraj, CEO of TII.

Outlook

The launch signals a tactical shift from scaling parameter counts toward architectural efficiency and practical deployment. By prioritising a compact, unified model, TII aims to support applications in manufacturing, robotics and enterprise operations where compute-constrained hardware and low-latency inference matter. Combined with Abu Dhabi’s integrated approach to research, governance and commercialisation — “accelerat[ing] adoption while maintaining oversight and trust,” as TII describes it — Falcon Perception illustrates the UAE’s intent to compete in the global multimodal AI landscape with homegrown, deployable systems.