Qatar announces Fanar 2.0 Arabic AI model
Sovereign model triples in size with multi-modal capabilities
#Qatar #LLMs - Qatar Computing Research Institute (QCRI) at Doha-based research university Hamad Bin Khalifa University (HBKU) has announced Fanar 2.0, an enhanced version of Qatar’s sovereign Arabic large language model, at the second World Summit AI in Doha. The new model is 27 billion parameters, compared with the Fanar model released last year 2024, which was 9B. According to QCRI, Fanar 2.0 family of models have multi-modal capabilities, being able to process text, images, speech-to-text, text-to-speech, audio understanding and Arabic poetry generation. The model operates within a fully closed ecosystem to guarantee data privacy and protection, whilst delivering deep understanding of Arabic dialects and cultural terminology.
SO WHAT? - Fanar 2.0’s emphasis on sovereign AI infrastructure addresses growing concerns among Arab governments and enterprises about data privacy when deploying international AI models that may share or circulate sensitive information beyond the user. By operating within a fully closed ecosystem, the platform enables organisations to leverage advanced AI capabilities without exposing confidential data to external risks, a critical requirement for government agencies and enterprises handling sensitive information. The threefold increase in model size combined with expanded multi-modal capabilities positions Fanar 2.0 as a comprehensive national AI solution that respects cultural norms and societal values whilst delivering competitive technical performance across text, speech and image generation applications.
Here are some ley points about the announcement:
Qatar Computing Research Institute (QCRI) at Doha-based research university Hamad Bin Khalifa University (HBKU) announced Fanar 2.0, the enhanced 27 billion parameter second version of Qatar’s sovereign Arabic large language model. The announcement was made during the opening of the second World Summit AI in Doha on 9 December 2025.
The new multimodal Fanar 2.0 is three times the weight of the original 9-billion parameter that was released in December 2024. Meanwhile, QCRI Executive Director Dr Ahmed Elmagarmid confirmed during the Summit that work has already commenced on Fanar 3.0 for release in December 2026.
In common with its predecessor, Fanar 2.0 offers multi-modal capabilities via a family of models including text processing, image generation, speech-to-text conversion, text-to-speech output, audio understanding and Arabic poetry generation. The model was trained to reflect Arabic cultural values and cope with Arabic language dialect diversity.
According to QCRI, the sovereign model operates within a fully closed ecosystem that allows organisations to run AI applications without sharing or circulating data beyond the user environment, guaranteeing the highest levels of privacy and data protection for government agencies and enterprises handling sensitive information.
Fanar is built using on a massive dataset containing more than 300 billion words and over one trillion Arabic phonetic segments, providing exceptional natural language processing capabilities with high precision across diverse Arabic dialects and cultural terminology aligned with local traditions and identity. No details have been shared regarding the differences between the data used or Fanar 1.0 and the corpus used for Fanar 2.0.
Fanar 2.0 will be made available to users through APIs and chat interface, enabling widespread adoption whilst maintaining the closed environment that ensures data security and cultural appropriateness for Arab organisations and individual users across the region.
Qatar Computing Research Institute employs more than 100 engineers and researchers working across various AI disciplines including AI-driven cybersecurity, data protection, secure language models, Arabic language processing, translation, speech technologies, and educational and healthcare applications supporting the platform’s development.
ZOOM OUT - Qatar launched its first sovereign large language model Fanar in December 2024 at the inaugural World Summit AI in Doha Developed by Qatar Computing Research Institute at Hamad Bin Khalifa University with sponsorship from the Ministry of Communications and Information Technology, the project brought together Qatar University, Qatar National Library, the Ministry of Endowment and Islamic Affairs, Al Jazeera and the Arab Center for Research and Policy Studies, with Google Cloud as technology provider. The original model was trained on 1.3 trillion tokens comprising 40% Arabic, 50% English and 10% code, with special curation of Qatari heritage, culture and dialect data.
[Written and edited with the assistance of AI]
LINKS
Fanar LLM website (includes no Fanar 2.0 details as of 10-Dec-25)
Read about the first version of Fanar LLM:
Qatar launches Fanar sovereign large language model (Middle East AI News)