The fastest world class Malay AI ever built. #2 in Southeast Asia, #7 globally, Trained in just 30 days Our Malay Large Language Model (LLM) has officially ranked on the MalayMMLU global benchmark, outperforming some of the most advanced models in the world including ChatGPT 5

This progress shows how quickly Malaysian-built AI can learn, adapt, and compete at the highest level, in a market where speed, accuracy, and technical depth matter. #STEM #GlobalBenchmark #MadeInMalaysia #AIinSTEM #SovereignAI

For Malaysia, strong STEM-capable AI means better decision making, stronger local innovation, and technology that can truly support our growing digital economy.

STEM matters because it sits at the core of real-world AI adoption, powering industries like finance, healthcare, manufacturing, energy, and education.

Ending the year strong with a major milestone in one of the most critical fields in AI: STEM! In our latest MalayMMLU STEM benchmark, our model achieved a score of 83.71, moving up to #4 globally. This is a significant jump from #11 just two weeks ago. 🚀

Season’s greetings from Aras Integrasi. 🎅🏻🎅🏻 Wishing everyone a Merry Christmas filled with joy, peace, and meaningful moments with loved ones 🎄🎄

Whether it’s linguistic nuance or academic depth, we’re committed to building an AI that truly understands the Malaysian context. #MalayLLM #AI #Humanities #Benchmark #TechMalaysia

✅ Islamic Studies: Understanding faith, practices, and global impact. ✅ Quran and Sunnah: Analyzing primary Islamic teachings. ✅ Sports Science Knowledge: Grasping human health and exercise science.

Across 4,395 questions, from the MalayMMLU benchmark, our model demonstrated deep contextual knowledge in specialized areas:

🚀 Another day, another ‘A’! Our LLM model’s expertise goes beyond just language proficiency. Alongside our top score in Malay Language, we’ve also achieved a Grade A in Humanities with an 81.16% accuracy rate.

Thank you to everyone who joined us and let’s continue to innovate and master the tools of tomorrow! #AIinEducation #AgenticAI #SovereignAI #DGXSpark

During the event, Dr. Shaggy led an in-depth session on Agentic & Sovereign AI. His expertise provided participants with critical insights into how autonomous AI agents can be developed and managed to ensure data sovereignty and technical excellence in university systems.

This initiative reinforces our commitment to driving meaningful digital transformation through both expert knowledge and practical exposure. We are especially proud to highlight the contribution of our own staff member, Dr. Shageenderan Sapai, who serves as our Lead of AI.

This provided participants with valuable hands-on experience, allowing them to explore real-world AI workloads and the practical applications of Agentic AI within a university environment.

This knowledge-sharing session focused on how intelligent AI agents are transforming university systems & application development. To bring these concepts to life, we brought two (2) DGX SPARK units to the event.

We have successfully concluded an ICT Webinar on Mastering Agentic AI for Universities' System & Application Development organised by UiTM's digital department.

It's more than just code! Today we break down the four essential elements needed to build effective, safe, and localized AI models. It starts with data and ends with ethics! Hit play to understand the foundation of safe and trustworthy AI! 🧠💻🇲🇾 #EffectiveAI #SovereignAI

It's more than just code! Today we break down the four essential elements needed to build effective, safe, and localized AI models. It starts with data and ends with ethics! Hit play to understand the foundation of safe and trustworthy AI! 🧠💻🇲🇾 #EffectiveAI #SovereignAI

As we continue improving our models, benchmarks like MalayMMLU are highly relevant to track our progress and show how far Malaysian-built AI has come especially within a specific linguistic and cultural context. Stay tuned, as our new benchmark is coming very soon!

MalayMMLU is important because it allows AI models to be tested within the local Malaysian context, while also being compared against top global models like GPT-4o and Llama 3.1. This makes it a crucial tool in supporting Malaysia’s long-term goal of AI sovereignty.

The MalayMMLU was first authored by Universiti Malaya (UM) and is accepted at EMNLP 2024, one of the world’s leading NLP conferences, giving it strong global credibility.