Saudi HUMAIN Launches Arabic LLM Built on MiniMax: what are the key facts?

HUMAIN released a 428-billion-parameter Arabic model, officially confirming it was developed based on MiniMax-M3.

Saudi HUMAIN Launches Arabic LLM Built on MiniMax: what are the key facts?

The model was trained on over 1 trillion Arabic tokens and achieved an average score of 89.37% across seven benchmarks.

Saudi HUMAIN Launches Arabic LLM Built on MiniMax: what are the key facts?

Chinese open-source models account for 41% of Hugging Face downloads, with Qwen derivatives exceeding 113,000.

Saudi HUMAIN Launches Arabic LLM Built on MiniMax: what are the key facts?

Institutions in Japan (Rakuten), Singapore (AI Singapore), and other nations are localizing development using Chinese model architectures.

Physix Frontier · News Briefing Card (enbrief · Sep 6, 2026)

Saudi HUMAIN Launches Arabic LLM Built on MiniMax

KEY FACTS

HUMAIN released a 428-billion-parameter Arabic model, officially confirming it was developed based on MiniMax-M3.
The model was trained on over 1 trillion Arabic tokens and achieved an average score of 89.37% across seven benchmarks.
Chinese open-source models account for 41% of Hugging Face downloads, with Qwen derivatives exceeding 113,000.
Institutions in Japan (Rakuten), Singapore (AI Singapore), and other nations are localizing development using Chinese model architectures.

KEY DATA

428BTotal Parameters

23BActive Parameters

1T+Training Tokens

41%HF Download Share

PHYSIX OBSERVATION

Chinese open-source models are shifting from product exports to infrastructure exports, becoming the foundational metric for global AI R&D. The Saudi case demonstrates that fine-tuning Chinese bases to build sovereign AI is now a trend, lowering the barrier to pre-training from scratch while reshaping global compute and data allocation, forcing closed-source giants to reassess their pricing and openness strategies.

Source: enbrief original report ↗