UAE releases 32B-parameter AI model that matches OpenAI performance

The United Arab Emirates has released K2 Think, an open-source AI reasoning model that matches the performance of much larger systems from OpenAI and DeepSeek while using just 32 billion parameters compared to their 200+ billion. Developed by researchers at Mohamed bin Zayed University of Artificial Intelligence in Abu Dhabi and backed by tech conglomerate G42, this represents one of the strongest indicators yet that the UAE’s massive AI investments are yielding competitive results in the global race for AI supremacy.

What you should know: K2 Think is specifically designed for advanced reasoning tasks rather than being a complete large language model, allowing it to deliberate through complex problems with remarkable efficiency.

• The model combines several cutting-edge innovations including fine-tuning on long reasoning sequences, agentic planning that breaks down problems multiple ways, and reinforcement learning focused on verifiably correct answers.
• G42 is running the model on Cerebras chips as an alternative to Nvidia hardware, demonstrating optimized performance on non-traditional AI infrastructure.
• MBZUAI has open-sourced the model and published detailed technical documentation explaining how the various innovations were integrated.

The big picture: This release positions the UAE as a serious contender in AI development alongside the United States and China, part of a broader trend of wealthy smaller nations developing “sovereign” AI models.

• The UAE has invested billions to establish itself as a strategically important research hub, including setting up Silicon Valley outposts and distancing itself from China to gain access to US silicon needed for frontier model training.
• Other Middle Eastern nations like Saudi Arabia are making similar heavy investments in AI infrastructure and research.

Why this matters: The achievement demonstrates that smaller, more efficient models can rival the performance of resource-intensive systems, potentially reshaping how the industry approaches AI development.

• “How to make a smaller model function as well as a more powerful one—that’s a lesson to learn, if other people want to learn from us,” said Eric Xing, MBZUAI’s president and lead AI researcher.
• This efficiency breakthrough could democratize access to advanced AI reasoning capabilities for organizations without massive computational resources.

Key technical details: The model’s development required significant but manageable computational resources compared to frontier models.

• Training used several thousand GPUs, with the final training run involving 200-300 chips.
• Plans are underway to incorporate K2 Think into a complete large language model in the coming months.

What they’re saying: Industry leaders see this as validation of the UAE’s strategic approach to AI development.

• “This is a technical innovation or, in my opinion, a disruption,” Xing told WIRED, emphasizing the effective combination of recent technical advances.
• “By proving that smaller, more resourceful models can rival the largest systems, this achievement shows how Abu Dhabi is shaping the next wave of global innovation,” said Peng Xiao, CEO of G42 and MBZUAI board member.

Geopolitical context: The release comes as President Trump traveled to the region in May to announce numerous AI deals involving US tech companies, highlighting the Middle East’s growing importance in the global AI landscape.