AI Insight

Falcon and Jais: How the UAE Is Building Its Own AI Model Ecosystem

Falcon and Jais represent two important pillars of the UAE’s homegrown AI ecosystem, combining efficient open models with Arabic-first intelligence.

AIU.AE Editorial23 Jul 2026

Falcon and Jais: How the UAE Is Building Its Own AI Model Ecosystem cover

The UAE is not waiting for global AI models to understand the region. It is building its own.

Most conversations about large language models are dominated by systems developed in the United States and China. However, the UAE has been steadily developing an increasingly capable homegrown AI ecosystem, led by model families such as Falcon and Jais.

These models are not identical, nor are they simply competing versions of the same technology. They approach the region’s AI requirements from different directions.

Falcon has developed into a broad family of efficient and openly available AI models spanning language, reasoning, Arabic intelligence and multimodal understanding. Jais has concentrated heavily on building advanced Arabic-first intelligence that understands the linguistic and cultural complexity of the Arab world.

Falcon: Building Powerful AI Through Efficiency

Falcon is developed by Abu Dhabi’s Technology Innovation Institute, the applied research pillar of the Advanced Technology Research Council.

The Falcon programme has become known for developing open AI models that aim to combine strong performance with practical deployment. Rather than relying only on increasingly large parameter counts, its recent models focus on architecture, efficiency, speed and lower infrastructure requirements.

One of the most important recent releases is Falcon H1R 7B, a compact seven-billion-parameter reasoning model. The model uses a hybrid Transformer–Mamba architecture designed to improve reasoning performance while reducing memory and computing requirements.

According to TII, Falcon H1R 7B demonstrated competitive performance against several significantly larger open models across mathematics, coding and general reasoning benchmarks. The model is openly available under the Falcon TII licence, allowing researchers and developers to experiment with and build upon it.

This development reflects an important shift in the AI industry: advanced intelligence does not always require the largest possible model. Stronger architecture and more efficient training can allow smaller models to deliver competitive results while being easier and less expensive to deploy.

Falcon Expands Into Arabic and Multimodal AI

The Falcon family is no longer focused only on general-purpose language models.

Falcon-H1 Arabic was developed specifically for Arabic language understanding. Its training covers Modern Standard Arabic, regional dialects and culturally grounded content, helping it address many of the limitations that Arabic speakers experience when using predominantly English-trained AI systems.

TII has also introduced Falcon Perception, a compact multimodal model that combines vision and language. It is designed to help machines see, read and understand information from the physical world, with potential applications in document processing, manufacturing, robotics and other real-world environments.

Together, these developments show Falcon evolving from a single language model into a broader UAE-developed AI model ecosystem.

Jais: Arabic Intelligence by Design

While Falcon covers a wide range of AI capabilities, Jais places Arabic language and regional context at the centre of its design.

The original Jais model was developed through a collaboration involving Inception, Mohamed bin Zayed University of Artificial Intelligence and Cerebras Systems. It was trained using large Arabic and English datasets on the Condor Galaxy AI supercomputing infrastructure.

Jais was created to understand more than direct Arabic translations. It was designed to handle Modern Standard Arabic, regional dialects, linguistic nuances and the cultural context behind the language.

The model family has since expanded through Jais 2, with open-weight 8-billion and 70-billion-parameter variants. According to Cerebras, Jais 2 uses an expanded Arabic-first dataset, an updated architecture and improved fine-tuning and alignment processes.

The 70-billion-parameter model is designed to provide strong Arabic performance while retaining multilingual reasoning and comprehension capabilities. This makes it relevant to governments, enterprises, researchers and developers building AI services for Arabic-speaking users.

Why Falcon and Jais Matter to the UAE

Falcon and Jais represent more than two families of large language models.

They strengthen the UAE’s ability to develop artificial intelligence that reflects the region’s own languages, culture, institutions and economic priorities.

Homegrown AI capabilities can support:

Arabic-first digital services that understand formal language and regional dialects.
Government and enterprise AI designed around local requirements.
Research and innovation through models that developers can access and build upon.
More efficient AI deployment across data centres, enterprise infrastructure and edge devices.
Greater technological resilience by reducing complete dependence on external model providers.

This does not mean the UAE must isolate itself from the global AI ecosystem. Sovereign capability is about having meaningful choices: the ability to use international technology while also possessing the expertise, infrastructure and models required to build locally.

Two Complementary Pillars of UAE AI

Falcon demonstrates how efficient architecture, open development and a growing range of model capabilities can make advanced AI more accessible.

Jais demonstrates why language, dialect and cultural context must be considered from the beginning rather than added as an afterthought.

Together, they show that the UAE is moving beyond being only a consumer or investor in artificial intelligence. It is developing the research, computing infrastructure and model-building expertise required to participate directly in the global AI industry.

The next phase of AI leadership will not be defined only by who adopts artificial intelligence fastest. It will also be defined by who can build, control, adapt and continuously improve it.

The UAE is positioning itself to do exactly that.