01 · Our Core Offerings
From Vision to Reality
LLM Training as a Service (TaaS)
In partnership with Phison’s aiDAPTIV+ and SNS, we offer scalable LLM training as a service, helping businesses unlock the power of AI for their own needs, from tailored language models to data management solutions.
LLM Gig Economy (Curatorship)
Malaysia’s first LLM Gig Economy Platform empowers Malaysians to participate in our AI’s growth through curated contributions, ensuring high-quality data inputs that keep Merdeka LLM relevant, accurate, and reliable.
AI Sovereignty as a Service
We empower enterprises and governments to achieve AI sovereignty with secure infrastructure, skilled talent, custom data solutions, and seamless deployment, keeping organisations in control of their AI systems and data.
02 · Why It's Crucial
The Importance of Sovereignty AI
Protecting National Interests
When Malaysia runs its own models, sensitive data stays in the country and out of foreign control.
Economic Independence
Job Creation: supports the gig economy, letting Malaysians contribute directly to AI development while earning income.
Innovation and Growth: local AI development fosters innovation and new business opportunities.
Reduced Dependency: less reliance on foreign technology, more self-reliance.
Data Ownership
Resource Control: Malaysians should control their own data, ensuring it’s used ethically and for the country’s benefit.
Economic Value: owning and managing data locally unlocks significant economic value across sectors.
03 · Intelligence Density
Powerful Intelligence, Smaller Footprint.
27B
parameters, compact enough to host on-premise
84.9%
on MalayMMLU, the highest of the Malaysian-built models
3.1
MalayMMLU points per billion parameters, the highest intelligence density in our comparison
1 GPU
to host it: a single 80 GB GPU holds the weights
Intelligence density: model size vs MalayMMLU accuracy
Higher and further left means more accuracy per parameter. Only models whose developers publish a parameter count. Hover a point for details.
MalayMMLU accuracy (%)
Fits one 80 GB GPU
(up to 40B)
(up to 40B)
90
85
80
75
70
10B
100B
1T
{{ pt.name }}
{{ pt.meta }}
{{ pt.tipTitle }}
{{ pt.tipSub }}
Total parameters (log scale)
Student data stays in school
Questions, answers and student work never leave the school’s own server: sovereign by design, and simpler for PDPA compliance.
Modest hardware
One GPU server is enough, so a school or district can own its AI instead of renting it.
No dependence on outside clouds
It runs on the school network, so lessons don’t stop when the internet or an overseas provider does.