Saudi Arabia just picked a side in the AI race, and it isn’t America’s. At LEAP in Riyadh, HUMAIN — the AI company owned by the kingdom’s $900B sovereign fund PIF — unveiled humain-m3, a 428B-parameter MoE model. The base? China’s MiniMax-M3.
What humain-m3 actually is
An Arabic-first frontier LLM. HUMAIN commissioned it, MiniMax built it: take the M3 lineage, continue pre-training on over 1 trillion tokens of native Arabic content. HUMAIN claims the highest average score across 7 public Arabic benchmarks, beating every frontier model tested. Self-reported for now — no independent numbers yet.
How to try it
The research preview is live on HUMAIN Node, HUMAIN’s API platform — full checkpoint, thinking and streaming on. Obvious use cases: Arabic chatbots, government services, anything where GPT-class models still stumble on dialects. Open weights land next month under the MiniMax Community License.
Why it matters
Sovereign AI was supposed to be America’s export product. Instead, a US ally built its national model on Chinese open weights — because you can’t fine-tune GPT-5 on a trillion Arabic tokens, but you can rebuild MiniMax. MiniMax’s stock jumped on the news. That’s the scoreboard.
You Might Also Like
- Minimax m2 5 Just Dropped and Open Source ai Will Never be the Same
- Mcp2cli the Tool That Cuts mcp Token Costs by 99 Just hit Hacker News
- Minimax m2 7 Scores 56 on swe pro and it Helped Build Itself
- Cursor Composer 2 Takes on Anthropic and Openai With a 0 50 m Token Coding Model and the Benchmarks Back it up
- 397 Billion Parameters on a 48gb Macbook Flash moe Turns Apples 2023 Research Into Reality
Top AI Product 
Leave a comment