The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) has significantly broadened its global footprint by launching its Institute of Foundation Models (IFM), a multi-site initiative featuring a new Silicon Valley lab in Sunnyvale, California. This lab complements existing IFM facilities in Paris and Abu Dhabi, marking a strategic expansion into one of the world’s most dynamic AI ecosystems.
Yesterday’s launch event at the Computer History Museum in Mountain View officially introduced the third node of MBZUAI’s growing global research network, connecting the university with California’s vibrant AI researchers, startups, and technology companies.
For the UAE, this expansion aligns with long-term economic diversification strategies aimed at strengthening knowledge-based sectors through cutting-edge AI technologies like advanced foundational models.
“Today’s launch of the IFM is an important step for the collaboration and global development of groundbreaking AI foundation models,” said Professor Eric Xing, MBZUAI President and University Professor. “Our expansion into Silicon Valley opens avenues for knowledge exchange and access to a deep talent pool that excels in turning research into real-world applications.”
PAN: A New World Model for Complex AI Simulations
Central to MBZUAI’s IFM demonstration was PAN, an advanced world model capable of simulating unlimited realities—from basic physics to intricate agent-based scenarios. Unlike prior AI systems mainly focused on text, audio, or images, PAN integrates multimodal inputs such as speech, video, spatial data, and physical actions to enable sophisticated reasoning, planning, and decision-making.
PAN’s hierarchical architecture supports multi-level reasoning and real-time simulation, offering accuracy and scalability for complex tasks like autonomous driving and robotics. Its companion system, PAN-Agent, showcases capabilities in multimodal reasoning tasks including mathematics and programming within dynamic simulated environments.
Advancing Foundational Models: K2 and JAIS
The IFM is also developing two flagship AI systems with global impact:
- K2: An advanced model focused on enhanced reasoning with sustained performance, with an upcoming update for the K2-65B version.
- JAIS: The world’s most advanced Arabic large-language model, which IFM aims to improve with broader language support and deeper contextual understanding to preserve and promote cultural heritage.
Transparency and Open Development at the Core
MBZUAI leads the industry in transparency by openly sharing not only AI models but entire development processes. The LLM360 initiative offers researchers access to training code, datasets, and model checkpoints, fostering open collaboration while maintaining rigorous safeguards such as international advisory boards and peer reviews to ensure research integrity.
IFM combines start-up agility with institutional resources, featuring dedicated teams for architecture, training, evaluation, and security, positioning MBZUAI at the forefront of ethical and pioneering AI research worldwide.
