Saudi Arabia’s Public Investment Fund-backed artificial intelligence company HUMAIN announced HUMAIN M3 on September 3, an Arabic-language model commissioned from Chinese AI developer MiniMax. Built on the MiniMax-M3 model family, it is available through HUMAIN Node as a research and evaluation preview, offering a concrete example of Chinese foundation-model technology supporting overseas AI development.
According to HUMAIN, the model uses a 428-billion-parameter mixture-of-experts architecture, activating 23 billion parameters per token, and underwent further pretraining on more than one trillion tokens of Arabic-native content. It supports text, image and video understanding, alongside tool use and computer-operation capabilities. HUMAIN’s evaluation recorded an equal-weighted average of 89.37% across seven public Arabic benchmarks, with the highest scores in five among the models compared. These are company-reported results.
HUMAIN is targeting October 2026 for the release of model weights under the MiniMax Community License, following safety training and alignment. The current offering remains an experimental preview rather than a production release.
MiniMax introduced M3 in June 2026, combining coding and agent capabilities with a one-million-token context window and native multimodal training. Its weights are publicly available under a community license that includes conditions on commercial use, making “open-weight” a more precise description than unrestricted open source.
The project adds to MiniMax’s international expansion. According to reporting on its first-half 2026 results, revenue from outside mainland China reached approximately USD 70.83 million, accounting for 60.8% of total revenue. The HUMAIN announcement did not disclose the engagement’s financial terms.
There is an earlier regional precedent. In October 2025, AI Singapore announced Qwen-SEA-LION-v4-32B-IT, built on Alibaba’s Qwen3-32B and adapted for Southeast Asian languages and cultural contexts. That example also rules out presenting HUMAIN M3 as the first national AI initiative to use a Chinese model foundation.
Together, the projects suggest an expanding overseas role for Chinese AI developers through model adaptation and development partnerships. For recipient countries, an existing foundation can reduce the work required to build local-language capabilities. The commercial opportunity lies in combining that foundation with local data, evaluation and deployment expertise; lasting influence will depend on licensing, performance and continued adoption.
Contact Us:
Xingyiran@iyiou.com
