Saudi Arabia's national artificial intelligence company HUMAIN says its full AI stack will be available in the Kingdom on 15 December, built around a 3 trillion parameter open-weight model that its chief executive describes as the largest of its kind in the world.

Chief executive Tareq Amin made the announcement on Wednesday evening at the company's offices in Majdoul Tower in Riyadh, where HUMAIN also unveiled HUMAIN+, an applied AI education and training platform due to begin operating in the first quarter of 2027. Tbreak, reporting the account first published by Arab News, said Amin gave the date plainly: “The complete stack, December 15, is not far away.”

The model itself remains unnamed. Amin said it would be the world's largest open-weight model, with 3 trillion parameters, and claimed it would cut cost and token use by 88 percent compared with models available outside Saudi Arabia. No model name, benchmark results, pricing or access route were published alongside the announcement, and the claims have not been independently verified.

What the December stack is supposed to include

HUMAIN, which is owned by Saudi Arabia's sovereign wealth fund, the Public Investment Fund, was set up in May 2025 to cover data centres, cloud platforms and AI models. The company says the December launch will bring together its AI software, cloud inference and infrastructure, all built, deployed and operated inside Saudi Arabia.

That framing is central to the pitch. Rather than renting intelligence from abroad, the Kingdom wants a domestic stack it controls, from the data centres up to the model itself. Amin disclosed that the development work behind the stack took about 15 months, a fast turnaround for a company that has existed for about 18 months.

The 88 percent figure is the claim that will face the closest scrutiny. Amin did not say which models he was comparing his system with, and Middle East AI News noted that no independent benchmark has been published. Until the weights and performance figures appear, the number is a company assertion rather than a demonstrated result.

HUMAIN+ and the training push

Alongside the model, HUMAIN inaugurated HUMAIN+, a Riyadh AI training centre built with VUZ, a Riyadh-founded immersive video technology company. The centre is expected to offer about 1,500 learner places a year from early 2027, with its operational launch now set for the first quarter of next year.

The training platform points to a second strand of the strategy: building a domestic workforce able to use the infrastructure, not only buying the hardware. Saudi Arabia has set out plans for 3 gigawatts of AI data-centre capacity by 2030, and HUMAIN sits at the centre of that ambition.

Why the launch date matters

15 December is now the first real test of Amin's promises. If HUMAIN publishes the model's weights, as an open-weight release implies, outside developers will be able to examine its size, its training and its performance for themselves. If the release is limited to Saudi Arabia, or arrives without benchmarks, the scepticism that greeted the announcement will harden.

There is recent precedent for caution. In September, HUMAIN unveiled humain-m3, a 428-billion-parameter Arabic-language model that the company later clarified was commissioned from the Chinese lab MiniMax and further trained on more than one trillion Arabic tokens. That model entered a limited research preview rather than a full public release.

The new model is a far bigger claim, and the stakes are higher. A 3 trillion parameter open-weight system would be the largest openly available model anywhere, if the figure and the release both hold. For now, the industry has a date, a number and a promise — and a month to wait before any of them can be checked.