HUMAIN promises 3-trillion-parameter model for December 15

HUMAIN, the artificial intelligence company of Saudi Arabia's sovereign wealth fund, the Public Investment Fund (PIF), has announced that its full AI stack will be available on December 15, 2026, including a 3-trillion-parameter open-weight model that CEO Tareq Amin describes as the largest in the world. The announcement was made on Wednesday evening, November 11, at the company's offices in Majdoul Tower, Riyadh, in front of journalists, AI experts and prospective HUMAIN+ participants.

The announcement

According to Amin, the 3-trillion-parameter model and the infrastructure arrive together in December. “By December, mark my words … the world's largest open-weight model, 3 trillion parameters model, will be available in Saudi (Arabia), nobody in the planet has this,” the executive said, according to Arab News.

Amin also stressed the sovereign character of the project: the software, the cloud inference and the infrastructure were built and are operated in Saudi Arabia. “Sovereign infrastructure, deployed here, ready to be consumed by any builder, enterprise, government without any control claim that leaves the country,” he said.

The executive said building the stack took 15 months and that the goal is to offer affordable computing to entrepreneurs, startups and students.

What has not been proven yet

The strongest claims in the announcement are, for now, statements by the CEO, without independent verification: the model still has no disclosed name, there are no public benchmarks, and no price or access route has been announced. Amin also claims the 3-trillion-parameter model will cost 88% less and consume 88% fewer tokens than models available outside Saudi Arabia — without saying which models the comparison is against.

December 15 will be the first test of these promises: whether HUMAIN publishes the model's weights and performance figures, and on what terms developers outside the country can access it.

HUMAIN+ arrives in the first quarter of 2027

At the same event, HUMAIN confirmed the schedule for HUMAIN+, its AI education and training platform, with an operational launch in the first quarter of 2027. The Riyadh campus was designed for about 1,500 learners a year, with GPU-backed Builder Labs and spaces for demonstrations and project showcases.

The platform was developed with VUZ, a Riyadh-founded company specializing in patented 360-degree video technology, and will combine the HUMAIN Voice and HUMAIN Create products. According to Amin, the idea grew out of dissatisfaction with existing training methods: “What if we create the Netflix of AI?”

Context: Saudi Arabia's sovereign bet

Created in May 2025 by the PIF, HUMAIN is the pillar of the Saudi strategy to become a global AI hub. In October, the company expanded its partnership with NVIDIA to up to 84 MW of AI computing, equivalent to about 27,000 GPUs (read the previous Geekiki article).

In September, HUMAIN had already launched humain-m3, an Arabic-language model with 428 billion parameters, developed from the MiniMax M3 base and released as a research preview — a precedent that reinforces caution: until the new model's weights and benchmarks are published, claims about it should be treated as company promises, not verified facts.

Geekiki will follow the arrival of the 3-trillion-parameter model on December 15 and the publication of the promised weights and benchmarks.

Image credit: official photo of the HUMAIN and VUZ partnership, via PR Newswire.

Sources

Transparency: This content was created, edited or reviewed with the assistance of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.

By Muse