Arabic Generative AI in Saudi Arabia: From ALLaM to HUMAIN Chat (2026)

Saudi Arabia is expanding generative artificial intelligence capabilities for the Arabic language. HUMAIN, backed by the Public Investment Fund, released the humain-m3 large language model at LEAP 2026, while King Abdullah University of Science and Technology (KAUST) established a dedicated Center of Excellence to advance deep learning research.

What is the new Arabic AI model launched in Saudi Arabia?

HUMAIN, a Saudi company owned by the Public Investment Fund, unveiled humain-m3 during the LEAP 2026 conference. Released as a research preview on the HUMAIN Node platform, the model focuses exclusively on Arabic, according to Al-Eqtisadiah newspaper.

MiniMax developed the system based on its MiniMax-M3 architecture. Built on a Mixture-of-Experts structure, humain-m3 contains 428 billion parameters and was pretrained on more than one trillion tokens of original Arabic content. It recorded the highest average performance among advanced models tested across seven general Arabic language benchmarks.

Tareq Amin, CEO of HUMAIN, noted that hundreds of millions of people speak Arabic, yet the language remains underrepresented in frontier AI. The release aims to strengthen Arabic's presence and make these capabilities available to developers, researchers, and innovators.

How does research infrastructure support generative AI development?

Development extends beyond commercial companies into academia. KAUST established its Center of Excellence for Generative AI on September 12, 2024, according to the university's website. Professor Bernard Ghanem leads the center alongside co-director Professor Juergen Schmidhuber.

The center focuses on three main pillars:

• Scientific research in deep learning technologies.
• Commercial innovation and translating research into applications.
• Developing the human talent and skills needed to work with the technology.

Professor Ghanem stated that Saudi Vision 2030 is transforming the Kingdom into an information-based economy, with AI serving as a major driver. The center's research aligns with four national priorities: human health, environmental sustainability, leadership in energy and industry, and future economies.

What initiatives has SDAIA launched for AI development?

The Saudi Data and Artificial Intelligence Authority (SDAIA) manages the national framework for organizing and developing the sector. Details of its data and AI projects are available on its initiatives page, according to SDAIA's official website.

Key steps and initiatives implemented by the authority include:

1. Launching the National AI Index to measure government entities' readiness to adopt the technology and improve institutional performance.
2. Introducing the One Million Saudis for AI (SMAI) initiative, offering free Arabic-language training programs and accredited certificates.
3. Launching the second version of the regulatory sandbox in 2025 to accelerate digital service innovation and support startups.
4. Signing strategic partnerships with global technology firms including AMD, Pure Storage, Data Direct Network, and Weka.io to develop specialized data centers and high-performance computing infrastructure.
5. Participating in international conferences such as LEAP 2025, where it signed seven memoranda of understanding and showcased applications like the Ainay project for detecting diabetic retinopathy.

Why is Arabic content both a challenge and an opportunity?

Despite being spoken by hundreds of millions, Arabic remains underrepresented in frontier AI models, as HUMAIN's CEO explained. The humain-m3 model addresses this gap through training on one trillion tokens of original Arabic content.

The HUMAIN Node platform gives companies and developers access to advanced AI models and the infrastructure required to scale reasoning capabilities, supporting application building, testing, and deployment. During the initial phase, the model is available strictly for research and evaluation, allowing researchers to test its abilities and provide feedback before the open release of model weights.

HUMAIN is building a full ecosystem covering models, applications, infrastructure, and data centers, alongside its ALLAM model family. The company expects to release the model weights under the MiniMax Community License after completing safety training and alignment processes, targeting the month following the launch announcement in September 2026.

Key figures

• humain-m3 parameters: 428 billion (Al-Eqtisadiah).
• Training data volume: More than one trillion tokens of original Arabic content (Al-Eqtisadiah).
• Benchmarks: Seven general Arabic language benchmarks, where the model achieved the highest average performance (Al-Eqtisadiah).
• Expected GDP contribution: Three percent of the Kingdom's GDP by 2030 (KAUST).
• Memoranda of understanding at LEAP 2025: Seven signed by SDAIA (SDAIA website).
• KAUST Center of Excellence launch date: September 12, 2024 (KAUST).

What should developers and organizations monitor next?

The open release of humain-m3 model weights will follow the completion of safety training, giving developers and researchers direct access to Arabic generative AI capabilities via HUMAIN Node. Organizations should also track outcomes from the second version of SDAIA's regulatory sandbox launched in 2025 to understand how private-sector AI applications will be governed. For institutions relying on Arabic content, early testing of these models offers a competitive advantage ahead of broader commercial availability.