8 Leading Qatari AI Models and Their Features and Strengths

Qatar is emerging as an important contributor to artificial intelligence research, particularly in Arabic-language AI, generative technologies and culturally aware machine learning. Through the Qatar Computing Research Institute (QCRI) at Hamad Bin Khalifa University (HBKU), the country has developed a growing collection of AI models designed to support language processing, translation, speech recognition, image generation and specialised knowledge applications.

A major contributor to this development is Fanar, Qatar’s Arabic-centric generative AI platform, developed with support from the Ministry of Communications and Information Technology. Its Fanar 2.0 ecosystem brings together specialised models designed for different applications.

Below is a list of eight notable Qatari-developed AI models and their features, capabilities and strengths. ¹

Summary of Qatari AI Models

No.	AI Model	Main Application	Key Strength
1	Fanar LLM	Text generation and reasoning	Arabic language intelligence
2	Fanar-Sadiq	Islamic knowledge assistance	Specialised Islamic content
3	Fanar-Oryx IVU	Image and video understanding	Arabic-aware visual analysis
4	Fanar-Oryx IG	AI image generation	Culturally relevant images
5	Fanar-Shaheen	Language translation	Arabic-English translation
6	Fanar-Aura	Speech recognition and generation	Arabic dialectal audio
7	Fanar-Diwan	Poetry generation	Classical Arabic poetry
8	FanarGuard	AI content moderation	Arabic cultural and safety alignment

1. Fanar LLM

Fanar LLM is Qatar’s flagship Arabic-centric large language model. Developed by QCRI, it is designed to understand and generate Arabic content while maintaining strong English-language capabilities.

Features:

• Arabic and English text generation.
• Support for Modern Standard Arabic and regional dialects.
• Question answering and summarisation.
• Advanced reasoning and tool calling.
• Support for research and educational applications.

Strengths: Its major advantage is its understanding of Arabic language structures, regional expressions and cultural context. Fanar 2.0 includes the Fanar-27B model, built on a Gemma 3 foundation. ²

2. Fanar-Sadiq

Fanar-Sadiq is a specialised AI model designed to address Islamic knowledge questions and provide relevant responses.

Features:

• Islamic content retrieval.
• Responses to Islamic knowledge questions.
• Specialised religious information processing.
• Integration into the wider Fanar AI ecosystem.

Strengths: Its specialised focus makes it useful for users seeking AI-assisted access to Islamic information rather than relying exclusively on general-purpose chatbots.

3. Fanar-Oryx IVU

Fanar-Oryx IVU is an image and video understanding model that analyses visual content and provides text-based explanations.

Features:

• Image recognition and interpretation.
• Video content understanding.
• Arabic visual question answering.
• Arabic calligraphy recognition.
• English and Arabic responses.

Strengths: Its ability to interpret culturally relevant visual material and Arabic calligraphy makes it valuable for education, research and visual content analysis. ³

4. Fanar-Oryx IG

Fanar-Oryx IG is the image-generation component of the Fanar Oryx family.

Features:

• Text-to-image generation.
• Arabic-oriented visual content creation.
• Culturally informed image generation.
• Creative visual production.

Strengths: It supports the creation of visual content that reflects Arabic cultural references and regional contexts, making it useful for designers, educators and content creators.

5. Fanar-Shaheen

Fanar-Shaheen is a specialised language translation model designed to facilitate communication between Arabic and English speakers.

Features:

• Arabic-to-English translation.
• English-to-Arabic translation.
• Support for Arabic dialects.
• Language understanding and contextual translation.

Strengths: Its focus on Arabic dialects helps address translation challenges involving regional expressions and conversational language.

6. Fanar-Aura

Fanar-Aura is the speech technology family within the Fanar ecosystem. It focuses on understanding and generating spoken language.

Features:

• Automatic speech recognition.
• Speech-to-text conversion.
• Text-to-speech capabilities.
• Arabic dialectal audio processing.
• Long-form audio transcription capabilities.

Strengths: It supports voice-based applications and makes Arabic audio processing more accessible across different dialects.

7. Fanar-Diwan

Fanar-Diwan is a specialised generative AI model focused on classical Arabic poetry.

Features:

• Arabic poetry generation.
• Classical poetic language production.
• Creative writing assistance.
• Arabic literary content generation.

Strengths: Its specialisation in Arabic poetry supports the preservation and exploration of Arabic literary traditions through artificial intelligence.

8. FanarGuard

FanarGuard is a specialised AI moderation model designed to improve safety and cultural alignment in Arabic-language AI systems.

Features:

• Arabic content moderation.
• Bilingual safety filtering.
• Identification of potentially harmful content.
• Culturally aware content assessment.

Strengths: It addresses the challenges of detecting unsafe or inappropriate content in Arabic, helping make AI applications more suitable for Arabic-speaking communities.

Qatar’s AI development programme demonstrates the growing importance of specialised artificial intelligence systems. Through Fanar and its associated models, the country is investing in Arabic language technologies, visual intelligence, speech processing, translation and cultural preservation.

These models offer opportunities for education, government services, research, creative industries and commercial applications. Their emphasis on Arabic language capabilities also helps address the underrepresentation of Arabic in global AI development.

Official platform: https://www.fanar.qa/

Research and model repository: https://huggingface.co/QCRI

Note: These are specialised models and components within Qatar’s broader AI ecosystem, rather than eight unrelated foundation-model providers.