Saudi Arabia's HUMAIN builds national AI platform on Chinese MiniMax model

Bloomberg reported that Saudi Arabia's state AI company HUMAIN has built a national AI platform on top of an open model from China's MiniMax Group. The report did not disclose the platform's exact name, its specific features, or which version of MiniMax's model it is based on.

To put it plainly: ten days earlier, on August 24, HUMAIN announced it would jointly develop an Arabic-specialized model with France's Mistral AI. This time, it has built a national AI platform using an open model from China's MiniMax. Drawing on open models from both the West and China at once suggests a strategy that avoids tying itself exclusively to any single country's technology bloc.

Who is MiniMax

The company trains its own models spanning text, images, video, and voice. Its video generation service goes by the name Hailuo, and its character chat app goes by Talkie — both under the same corporate roof. Because the branding is split across several names, people often mistake these for separate companies, but the open model HUMAIN adopted comes from the same company's technology lineage.

HUMAIN follows Mistral with MiniMax

HUMAIN's partnership with France's Mistral stood out for its scale. On August 24, the two companies announced a strategic collaboration worth several hundred million euros, covering AI infrastructure buildout, joint development of a frontier model with strong Arabic capabilities, and deployment of AI solutions across Saudi Arabia and the surrounding region. Early efforts, they said, would focus on cybersecurity and voice.

Just ten days after that announcement, word emerged that HUMAIN had also adopted an open model from China's MiniMax as the foundation for its national platform. Taken together, the two moves suggest HUMAIN's strategy is not to lock itself into any one technology camp. With Mistral, it entered a development partnership to build an Arabic model from scratch; with MiniMax, it chose to adopt an already-trained open model and layer it onto its own national platform.

Category	Mistral AI Partnership	MiniMax-based Platform
Announced	August 24, 2026	September 3, 2026 (per report)
Mode of cooperation	Joint development partnership	Adoption of open model
Scale	Several hundred million euros	Undisclosed
Focus areas	Cybersecurity, voice, Arabic frontier model	Broad national AI platform (details undisclosed)

A growing race among nations for control over AI technology

Bloomberg framed this case within a growing list of countries seeking greater control over AI technology. Adopting open models can save considerably on cost and time compared with training large models from scratch, making it an accessible path for countries building out their own data centers and regulatory frameworks. Saudi Arabia is among the countries that have ramped up investment in data centers and AI infrastructure through its sovereign wealth fund, with HUMAIN serving as the state's vehicle for carrying out that investment.

Editor's take

What stands out here is that HUMAIN isn't locking itself into a single technology camp. The typical image of a national AI sovereignty strategy is training a model from scratch with domestic talent. Instead, within the span of ten days, HUMAIN signed a joint development deal with a French company and adopted an open model from a Chinese company as a foundation. That looks like a choice that prioritizes speed of deployment and practical application over technological self-reliance.

Building national platforms on open models has become a trend that's hard to ignore. Just a few years ago, national AI projects leaned heavily toward training their own foundation models from the ground up. But as open-weight models have closed the performance gap with commercial ones, adopting rather than building has started to look faster and cheaper. Under this dynamic, expect more countries to focus not on building their own models, but on choosing which open model to adopt and adapting it to their own language and regulatory needs.

There's a lesson here for domestic companies and government agencies weighing their own AI sovereignty questions. If training a large model from scratch isn't feasible given the time and cost involved, adopting a trustworthy open model and adapting it to local language and regulatory conditions can be a realistic alternative. Still, questions about which country's technology underlies that model, and where the data actually flows, remain separate issues that need to be worked out.

It's likely that HUMAIN will disclose the platform's name and the scope of its services in more detail within the coming weeks.