Colibrì Lifts the Elephant: How a Tiny Program Runs Giant AI on a Laptop

A small bird is trying to lift an elephant. That is roughly the first impression one gets from reading the story of Colibrì, a software project that enables a budget-class computer to run a giant AI brain such as GLM-5.2.

The AI brain called GLM-5.2 is built from roughly 744 billion parameters — numbers learned during training that determine how the AI recognises patterns, understands questions and composes answers. Even after compression, its file size remains around 370 GB.

Remarkably, however, Colibrì can run it on a 12-core laptop with only 25 GB of RAM. It does not load the entire giant parameter set into the computer’s RAM at once. It only fetches the parts currently needed, while the rest stays stored on an SSD or NVMe drive and is called up only when its turn comes to work.

For now, Colibrì is indeed only carrying the “elephant” known as GLM-5.2. This AI brain uses an approach called Mixture-of-Experts (MoE). It can be likened to a company with 744,000 employees, yet each task only requires around 40,000 people whose expertise matches the job.

The remaining 704,000 need not crowd into the office; they can wait until their expertise is required. That is the gap Colibrì exploits. In the conventional way, running GLM-5.2 demands a computer with extraordinary memory or several top-class graphics cards that are certainly not priced for the budget market.

Colibrì chose a different path: what is currently working goes into RAM, while what is not yet needed waits on the SSD. The technical details are naturally more complicated. What is certain is that the Colibrì experiment proves the impossible can become possible through creativity and ingenuity in solving a problem.

The name Colibrì itself feels like a metaphor almost too perfect to be a coincidence. In Italian, colibrì means hummingbird, a tiny creature from the Americas with unusual flying abilities. It can hover, spin, dart, and even fly backwards.

Its small body requires a constant supply of energy from flower nectar. The hummingbird does not become mighty by growing its body like an eagle. It overcomes its limitations through agility and a different way of moving.

So it is with the software bearing its name. It does not confront a giant AI model with an equally giant computer. It moves between RAM, VRAM and the computer’s storage, taking what is needed, leaving what is not yet required, then shifting again.

It does not defeat weight with strength, but with movement. It does not match enormity with enormity, but with ingenuity.

Here lies Colibrì’s broader lesson: limitations do not always have to be answered by adding power; sometimes they instead force us to find new ways through creativity.

We live in an era fond of solving shortages through additions. Slow computer? Add more RAM. AI model doesn’t fit? Buy a GPU. Server not powerful enough? Buy a new server. Company short on capacity? Increase the budget. None of that is wrong, especially if there is money to buy it.

Colibrì was born precisely because its creator, Vincenzo, known as JustVugg, did not possess a machine considered worthy of running a model that large. He could simply have accepted his fate: a 744-billion-parameter model is indeed not for a laptop like mine. End of story.

But Vincenzo did not stay still. He posed a different question: which parts of such a huge model are truly needed at any given moment? From that question, a seemingly solid wall began to have a door.

In the history of technology, doors like that are often found by those who lack the money to buy a new wall. Limitation indeed has two faces: an excuse to give up, or the mother of creativity.

For creativity is often not the ability to make something more complicated, but rather to find a simple path unseen because we are too accustomed to the expensive one.

For us in Indonesia, this lesson feels close to home. Not every school, Islamic boarding school, university, library, law office, community health centre or research institute can afford an AI server worth hundreds of millions to billions of rupiah.